Search

Purpose: In patients with metastatic prostate cancer (mPC), ATM and BRCA2 mutations dictate differences in PARPi inhibitor response and other therapies. We interrogated the molecular features of ATM- and BRCA2-mutated mPC to explain the divergent clinical outcomes and inform future treatment decisions. Experimental design: We examined a novel set of 1,187 mPCs after excluding microsatellite-instable (MSI) tumors. We stratified these based on ATM (n = 88) or BRCA2 (n = 98) mutations. As control groups, mPCs with mutations in 12 other homologous recombination repair (HRR) genes were considered non-BRCA2/ATM HRR-deficient (HRDother, n = 193), whereas lack of any HRR mutations were considered HRR-proficient (HRP; n = 808). Gene expression analyses were performed using Limma. Real-world overall survival was determined from insurance claims data. Results: In noncastrate mPCs, only BRCA2-mutated mPCs exhibited worse clinical outcomes to AR-targeted therapies. In castrate mPCs, both ATM and BRCA2 mutations exhibited worse clinical outcomes to AR-targeted therapies. ATM-mutated mPCs had reduced TP53 mutations and harbored coamplification of 11q13 genes, including CCND1 and genes in the FGF family. BRCA2-mutated tumors showed elevated genomic loss-of-heterozygosity scores and were often tumor mutational burden high. BRCA2-mutated mPCs had upregulation of cell-cycle genes and were enriched in cell-cycle signaling programs. This was distinct from ATM-mutated tumors. Conclusions: Tumoral ATM and BRCA2 mutations are associated with differential clinical outcomes when patients are stratified by treatments, including hormonal or taxane therapies. ATM- and BRCA2-mutated tumors exhibited differences in co-occurring molecular features. These unique molecular features may inform therapeutic decisions and development of novel therapies.

The Defense Advanced Research Projects Agency (DARPA) OFFensive Swarm-Enabled Tactics (OFFSET) program seeks to develop swarms of up to 250 aerial- and ground-based platforms to aid small-unit infantry forces to accomplish missions in complex urban environments. Over the course of four years, the OFFSET performers regularly tested and improved the autonomy, hardware, communications infrastructure, and logistics necessary to enable such a vision. In this paper, we present the Immersive Interaction Interface (I3), a virtual reality interface designed by Smart Information Flow Technologies (SIFT), as part of the BBN/Raytheon-led Command and Control of Aggregate Swarm Tactics (CCAST) team. I3 served as the main control interface for the team throughout the program, and this paper discusses the primary innovative features of I3 as compared to prior swarm control interfaces, how I3 was used in the field during exercises, and lessons learned over the course of the program. The paper also presents numerical results from the final three field exercises of the OFFSET program, demonstrating how a single operator used I3 effectively to control large numbers of unmanned vehicles.

Control of robotic swarms through control over a leader(s) has become the dominant approach to supervisory control over these largely autonomous systems. Resilience in the face of attrition is one of the primary advantages attributed to swarms yet the presence of leader(s) makes them vulnerable to decapitation. Algorithms which allow a swarm to hide its leader are a promising solution. In prior work we found that using a graph neural network, GNN, a swarm could be trained to flock following a leader. An Adversary NN trained to identify that leader (naïve condition) performed substantially better than human observers. When the swarm was trained to hide its leader (deception conditions), however, the advantage reversed with humans outperforming the Adversary. This human advantage persisted even when the swarm and Adversary were jointly trained, allowing the Adversary to adapt to the swarm’s evolving strategies for hiding its leader. The present study investigates the robustness of human leader identification by testing identifications made in the presence of medium and high levels of visual clutter. Clutter degraded human performance to some extent but human accuracy in leader identification remained well above that of the Adversary in deception conditions. Human performance even approached that for an unhidden leader under joint training. This study confirms the robustness of the human superiority effect and argues for the inclusion of humans in AI systems which may confront learned deception.

Tumor mutation burden (muts/Mb) of ATM and BRCA2 mutant mPCs.

Comparisons of WTS profiles between ATMm and BRCA2m to the HRP group.

AR activities and NEPC signatures in ATM- and BRCA2- mutated tumors.

Overall survival of ATM, BRCA2 mutant, and HRDother mPCs.

Composition and rates of all detectable genomic features detectable through the Caris diagnostic platform.

Purpose:In patients with metastatic prostate cancer (mPC), ATM and BRCA2 mutations dictate differences in PARPi inhibitor response and other therapies. We interrogated the molecular features of ATM- and BRCA2-mutated mPC to explain the divergent clinical outcomes and inform future treatment decisions.Experimental design:We examined a novel set of 1,187 mPCs after excluding microsatellite-instable (MSI) tumors. We stratified these based on ATM (n = 88) or BRCA2 (n = 98) mutations. As control groups, mPCs with mutations in 12 other homologous recombination repair (HRR) genes were considered non-BRCA2/ATM HRR-deficient (HRDother, n = 193), whereas lack of any HRR mutations were considered HRR-proficient (HRP; n = 808). Gene expression analyses were performed using Limma. Real-world overall survival was determined from insurance claims data.Results:In noncastrate mPCs, only BRCA2-mutated mPCs exhibited worse clinical outcomes to AR-targeted therapies. In castrate mPCs, both ATM and BRCA2 mutations exhibited worse clinical outcomes to AR-targeted therapies. ATM-mutated mPCs had reduced TP53 mutations and harbored coamplification of 11q13 genes, including CCND1 and genes in the FGF family. BRCA2-mutated tumors showed elevated genomic loss-of-heterozygosity scores and were often tumor mutational burden high. BRCA2-mutated mPCs had upregulation of cell-cycle genes and were enriched in cell-cycle signaling programs. This was distinct from ATM-mutated tumors.Conclusions:Tumoral ATM and BRCA2 mutations are associated with differential clinical outcomes when patients are stratified by treatments, including hormonal or taxane therapies. ATM- and BRCA2-mutated tumors exhibited differences in co-occurring molecular features. These unique molecular features may inform therapeutic decisions and development of novel therapies.

PurposeUsing a real-world database with matched genomic-transcriptomic molecular data, we sought to characterize the distinct molecular correlates underlying clinical differences between young-onset pancreatic cancer (YOPC; MethodsWe analyzed matched whole-transcriptome and DNA sequencing data from 2430 patient samples (YOPC, n=292; AOPC, n=2138) from the Caris Life Sciences database (Phoenix, AZ). Immune deconvolution was performed using the quanTIseq pipeline. Overall survival (OS) data was obtained from insurance claims (n=4928); Kaplan-Meier estimates were calculated for age-and molecularly-defined cohorts. Significance was determined as FDR-correctedP-values (Q)ResultsYOPC patients had higher proportions of mismatch repair-deficient (dMMR)/microsatellite instability-high (MSI-H),BRCA2-mutant, andPALB2-mutant tumors compared with AOPC patients, but fewerSMAD4-, RNF43-, CDKN2A-, andSF3B1-mutant tumors. Notably, YOPC patients demonstrated significantly lower incidence ofKRASmutations compared with AOPC patients (81.3% vs. 90.9%;Q=0.004). In theKRAS-wildtype subset (n=227), YOPC tumors demonstrated fewerTP53mutations and were more likely driven byNRG1andMETfusions, whileBRAFfusions were exclusively observed in AOPC patients. Immune deconvolution revealed significant enrichment of natural killer (NK) cells, CD8+T cells, monocytes, and M2 macrophages in YOPC patients relative to AOPC patients, which corresponded with lower rates ofHLA-DPA1homozygosity. There was an association with improved OS in YOPC patients compared with AOPC patients withKRAS-wildtype tumors (median 16.2 [YOPC-KRASWT] vs. 10.6 [AOPC-KRASWT] months;P=0.008) but notKRAS-mutant tumors (P=0.084).ConclusionIn this large, real-world multi-omic characterization of age-stratified molecular differences in PDAC, YOPC is associated with a distinct molecular landscape that has prognostic and therapeutic implications.

Control of robotic swarms through control over a leader(s) has become the dominant approach to supervisory control over these largely autonomous systems. Resilience in the face of attrition is one of the primary advantages attributed to swarms yet the presence of leader(s) makes them vulnerable to decapitation. Algorithms which allow a swarm to hide its leader are a promising solution. We present a novel approach in which neural networks, NNs, trained in a graph neural network, GNN, replace conventional controllers making them more amenable to training. Swarms and an adversary intent of finding the leader were trained and tested in 4 phases: 1-swarm to follow leader, 2-adversary to recognize leader, 3-swarm to hide leader from adversary, and 4-swarm and adversary compete to hide and recognize the leader. While the NN adversary was more successful in identifying leaders without deception, humans did better in conditions in which the swarm was trained to hide its leader from the NN adversary. The study illustrates difficulties likely to emerge in arms races between machine learners and the potential role humans may play in moderating them.

This panel will discuss issues related to Manned-Unmanned Teaming (MUMT) technologies. Panelists were selected to represent diverse topics and each will provide a unique perspective on the MUMT challenge space. Joseph Lyons will frame the discussion and introduce the panelists. Each panelist will provide an overview of the MUMT research/applications they are involved in. Chris Miller will discuss an ongoing project looking at MUMT applications broadly across the enterprise and is seeking to identify the core systemic tenants of MUMT and metrics to gauge MUMT effectiveness. Jay Shively will discuss MUMT challenges in the context of UAS operations in the National aerospace. Nancy Cooke will discuss several MUMT research projects that emphasize teaming and associated research challenges. Col. Dan Javorsek will discuss recent MUMT programs at DARPA as well as where MUMT technologies can support Air Force applications. Phillip Walker will discuss the DARPA OFFSET program and human-swarm interactions, including human factor considerations of large swarm demonstration events.

BACKGROUND Older age is a poor prognostic factor for glioblastoma (GBM) patients. We tested whether the intrinsic molecular landscape of the tumor may contribute to this poor prognosis. METHODS In accordance with the 2021 WHO classification scheme, we included only isocitrate dehydrogenase (IDH) wild type GBM. Based on published literature, we defined older as age > 65. RNA expression, gene amplification, tumor mutational burden (TMB) and mutational profiles were analyzed in three unique datasets: Tempus (n = 1,410), Caris (n = 1,432), and TCGA (n = 557). Comparison were made between < 65 and ³ 65 year olds using Pearson’s Chi-squared tests, Fisher’s exact tests, or Wilcoxon rank-sum where appropriate. RESULTS From our evaluable gene sets, TERT promoter mutations were more prevalent in patients ³ 65 years old (Caris 82.64 vs 77.27%, p = 0.016; Tempus 58.0 vs 49.0%, p = 0.002). There were no significant differences in PDCD1, CD274, CD3E, TNFRSF18, CD40, CD8A, TNFRSF4, CTLA4, HAVCR2, TNFSF9, CD274, or CDKN2A; PDL-1 (by IHC); dMMR/MSI-H, TMB; CDK6 amplification, EGFR amplification, EGFR, EGFRvIII, EGFR fusions, MET fusions, PTEN, TP53, or NF-1. MGMT promoter methylation (Caris data) was more common in the older group (49.73 v 34.14%, p < 0.001). TGCA data demonstrated that gene expression, TMB, and methylation did not change significantly with age. Additionally, PCOLCE2 and SLC10A4 were differentially methylated, and missense mutations, of any type, were more common in the older group (p=0.006). CONCLUSION Despite worse survival outcomes for older patients with IDHwt GBM as compared to younger counterparts, the molecular landscape is similar at the genomic, transcriptomic and epigenomic levels. The key exception is TERT promoter mutations that are more common in older GBM patients. Poorer survival is therefore not likely to be attributable solely to intratumoral factors.

INTRODUCTION Leptomeningeal disease (LMD) is a challenging complication of high grade glioma (HGG) and critical questions remain unanswered regarding clinicopathologic risk factors, molecular associations, and optimal treatment. METHODS Patients with molecularly-profiled HGG (Caris Life Sciences; Phoenix, AZ) with LMD at two institutions were included. Medical records were reviewed for clinicopathological characteristics, treatment, and outcome. Kaplan-Meier estimates of patient survival were performed on censored data using Cox’s proportional hazard model. RESULTS 43 patients (male: 33, female: 10; median age: 56 years) were identified, comprising 41 grade 4 (glioblastoma: 38; gliosarcoma: 2; H3K27M diffuse midline glioma: 1) and 2 grade 3 tumors (astrocytoma: 1; pleomorphic xanthoastrocytoma: 1). LMD diagnosed at HGG diagnosis (n=18) versus recurrence (n=22) was associated with longer post-LMD survival [pLMD-OS: 15.3m vs. 4.8m, HR: 0.07, 95% CI: 0.02-0.29, p=0.0004] but similar overall survival [mOS: 15.3m vs. 12.3m; HR: 0.82; 95% CI: 0.36-1.85; p=0.63]. Pathology-diagnosed LMD (n=15) versus MRI-diagnosed LMD (n=26) was associated with longer post-LMD survival [pLMD-OS: 15.4m vs. 5.2m, HR: 14.9, 95% CI: 0.01-0.30, p=0.0004] but similar overall survival [mOS: 17.1m vs. 12.3m; HR: 0.66; 95% CI: 0.3-1.58; p=0.38]. Post-LMD survival was significantly prolonged for supratentorial (n=28) versus infratentorial/spinal (n=4) locations regardless of the diagnostic modality [pLMD-OS: 2.6m vs. 11.3m, HR: 14.4, 95% CI: 2.73-75.7, p=0.0017], and did not significantly differ between symptomatic (n=20) and asymptomatic (n=23) patients [pLMD-OS: 4.8m vs. 11.2m, HR: 1.75, 95% CI: 0.82-3.77, p=0.15). pTERT mutation (81%), EGFR amplification (43%), and MGMT methylation (33%) were prevalent but IDH1 mutation was rare (2.8%). Comparison with a separate glioblastoma cohort (n=1400) suggested more frequent amplification of CHIC2, MDM4, and KDR, higher mutation rates of RUNX1, APC, and RAD51C, colder tumor microenvironment (TME), and lower expression of immune checkpoint-related genes. CONCLUSIONS Clinicopathological characteristics affect post-LMD survival, and cohort comparison suggests molecular and TME differences in LMD-HGG tumors.

713 Background: The FOLH1 gene encodes prostate-specific membrane antigen (PSMA), a transmembrane glycoprotein that is highly expressed in prostate cancer cells and on endothelial cells in the neovasculature of solid tumors, including RCC. PSMA has been used as a target for diagnostic imaging and therapeutic radioligand therapy. We utilized a database of molecularly profiled RCC tumors to evaluate associations with FOLH1 expression. Methods: NextGen sequencing of DNA (592-gene/whole exome) and RNA (whole transcriptome) was performed for RCC patient specimens (n=1765) through Caris Life Sciences (Phoenix, AZ). FOLH1-High/Low expression were defined as ≥75th/

562 Background: Recent data suggest that patients with liver metastases (mets) are resistant to immune checkpoint inhibition (CPI) independent of historical biomarkers of CPI efficacy, raising the question of whether liver mets may be associated with an immunosuppressive tumor microenvironment (TME). We investigated the immune TME of hepatic and non-hepatic mets compared to primary tumors in advanced urothelial carcinoma (UC) tissue samples. Methods: NextGen sequencing (NGS) of DNA (592-gene/whole exome) and RNA (whole transcriptome) from UC tissue samples (N=4746) was performed at Caris Life Sciences (Phoenix, AZ). Immune cell infiltration was estimated by RNA expression deconvolution (MCP-counter). PD-L1 expression (SP142: immune cell stain ≥ 5%; 22c3: CPS ≥ 10) was assessed by immunohistochemistry (IHC). Deficient mismatch repair/high microsatellite instability (dMMR/MSI-H) was tested by IHC/NGS. Real-world overall survival (OS) information was obtained from insurance claims data and Kaplan-Meier estimates were calculated. Mann–Whitney U and X2/Fischer-Exact tests were applied where appropriate, with p-values adjusted for multiple comparisons (Benjamini-Hochberg). *P+ T and B cells (0.55* and 0.29-fold*) but increased monocytic lineage cells (1.23-fold*), while non-hepatic mets had increased CD8+ T, NK, and monocytic lineage cells (1.28*, 1.27*, 1.31-fold*) with no difference in B cells (1.05-fold). Hepatic mets had decreased expression of integrin LFA-1/ ITGAL (0.77-fold*), as well as hyaluronic acid (HA) receptor CD44 and synthase HAS2 (0.61 and 0.61-fold*), compared to primary tumors, whereas expression of these genes and LFA-1 ligand ICAM1 was increased in non-hepatic mets (1.08 to 1.30-fold*). Hepatic mets had increased expression of immunosuppressive cytokines CCL2 and CXCL2 (1.72* and 2.32-fold*) and decreased expression of pro-inflammatory cytokines CCL5 and CXCL10 (0.63* and 0.79-fold*) compared to primary tumors. PD-L1+ IHC was less frequent in hepatic mets compared to primary tumors and non-hepatic mets. TMB-High (≥10 mut/MB) and dMMR/MSI-H rates were similar across tumor sites. Hepatic mets (N=40) were associated with worse OS from the start of pembrolizumab compared to non-hepatic mets (N=177) (19.6 vs 4.4 months, HR 3.01, 95% CI 1.91-4.75, p

741 Background: While pancreatic adenocarcinoma (PDAC) remains a leading cause of cancer-related deaths, the highly aggressive PDAC subtype of undifferentiated sarcomatoid carcinoma (USC) remains poorly characterized as it comprises only 2-3% of all PDAC histology. Previous case reports suggest that immune checkpoint inhibitors could be a promising treatment strategy for USC, but the prevalence of established predictive biomarkers of response are largely unknown in this unique subpopulation. We identified PDAC USC patient samples from a large dataset and performed comprehensive genomic profiling to determine the prevalence of biomarkers associated with response to immunotherapy. Methods: PDAC USC patient samples (N=43) underwent central pathology review to confirm this diagnosis and were compared to non-USC PDAC patient samples (N=5562). Next-generation sequencing of DNA and RNA was performed at Caris Life Sciences (Phoenix, AZ). PD-L1 expression was tested by IHC (SP142; Positive (+): ≥ 2+, ≥%5). A combination of IHC, NGS, and fragment analysis was used to assess deficient mismatch repair/microsatellite instability high (dMMR/MSI). High tumor mutational burden (TMB-High) was defined as ≥10 mutations/MB. Immune cell fractions of the tumor microenvironment were estimated by RNA deconvolution analysis using quanTIseq. Chi-square tests with Bonferoni correction for multiple comparisons were used to determine significance. Results: Among PDAC USC samples, KRAS (86% USC vs 90% non-USC, p = 0.31, q = 1), TP53 (86% vs 77%, p = 0.16, q = 1), and CDKN2A (18% vs 23%, p = 0.45, q = 1) were the most commonly mutated genes with a similar prevalence compared to non-USC histologies; however, KRAS was amplified more frequently (7% USC vs 5%) with neutrophils (85% vs 57%, q = 0.03) or M2 macrophages (52% vs 28%, p = 0.006, q = 0.06). Conclusions: This work represents the largest molecular analysis of PDAC USC samples to date. Our analysis uncovered a different prevalence of amplified KRAS and PD-L1 expression in USC as compared to other PDAC histologies amidst an immune desert lacking lymphocytes in the USC tumor microenvironment. These findings provide evidence for further investigation into combination therapy of KRAS inhibitors with immune checkpoint inhibitors to target these immune-imbalanced microenvironment landscapes.

185 Background: Down syndrome (DS), a genetic disorder caused by trisomy of chr 21, is associated with a considerably lower risk for solid tumors and other angiogenesis related diseases. DSCR1 belongs to a family of evolutionary conserved protein-coding genes located on chr 21 and is highly upregulated in DS patients. Its product, calcipressin-1, has been shown to reduce cancer risk by suppressing angiogenesis. We previously reported that a germline polymorphism in DSCR1 was associated with time to recurrence in resected CRC and anti-VEGF treatment outcomes in metastatic CRC. Here, we analyzed the molecular landscape of CRC according to DSCR1 expression levels. Methods: 20,237 samples from CRC tested at Caris Life Sciences (Phoenix, AZ) with WTS (Illumina NovaSeq) and NextGen DNA sequencing (NextSeq, 592 Genes and NovaSEQ, WES) were analyzed. Top quartile transcripts per million (TPM) for DSCR1 expression were considered high (Q4) while bottom quartile low (Q1). Cell infiltration (CI) in the tumor microenvironment (TME) was estimated by RNA deconvolution analysis using QuantiSEQ. Interferon-gamma and T-cell inflamed signatures were also calculated from RNA data. X2 and Fisher-Exact tests were used and statistical significance was determined as P-value adjusted for multiple comparisons ( q < 0.05). Results: DSCR1 expression was higher in primary tumors than metastases (10.5 vs 8.1 median TPM, q < 0.05). No significant difference was observed in right- versus left-sided tumors, however rectal tumors showed the highest DSCR1 expression ( P < 0.05). Overall, high DSCR1 TPM were associated with TMB-high (11.3% vs 9.3%), dMMR/MSI-H (7.9% vs 5.9%), and PD-L1 (4.7% vs 3.1%) ( q < 0.01); the association with TMB-H was not significant in pMMR/MSS. DSCR1 high was associated with lower mutation rates of APC, KRAS, TP53 and amplification of FLT1/ FLT3, while higher mutation rates of KMT2C/D, BRAF, PTEN, RNF43, and RSPO3 fusions ( q < 0.0001). Gene set enrichment analysis showed that high DSCR1 expressing tumors were enriched in alterations of several pathways including hypoxia, apoptosis, DNA repair, KRAS signaling, inflammatory response, and oxidative stress-related pathways (all P < 0.05, FDR < 0.25). B cells, macrophages, neutrophils, NK cells, Tregs, cancer associated fibroblasts and endothelial cells were more abundant in the TME of tumors with high DSCR1 while myeloid dendritic cells were lower, regardless of MMR status (all q < 0.001). DSCR1 expression was associated with a higher T-cell inflamed signature and IFN score ( q < 0.05). Conclusions: This is the first and most extensive profiling study to investigate DSCR1 expression in CRC. Our data show a strong association between tumor DSCR1 gene expression and distinct molecular features and TME cell infiltration. These findings suggest that DSCR1 holds potential as a novel therapeutic target for CRC and may be an important player in TME modulation.

592 Background: Tumor-agnostic approvals of immune checkpoint inhibitors (ICI) include deficient mismatch repair/microsatellite instability high (dMMR/MSI-H) and high tumor mutational burden (TMB) while in cancer like metastatic gastroesophageal cancers, ICI use has been predicated on PD-L1 expression. ICI are increasingly used for metastatic HCC, but without required markers. We aimed to examine the genomic landscape of HCC in context of PD-L1 expression, and to determine clinical responses to ICI in this setting. Methods: Next-generation sequencing of DNA (592 or WES) and RNA (WTS) was tested at Caris Life Sciences (Phoenix, AZ). PD-L1 expression was tested by IHC (SP142) and compared as high (2+,5%), low (1-2,1%) and negative (0). dMMR/MSI-H was tested by IHC/NGS and TMB-High was defined as ³10 mutations/MB. QuantiSEQ was used to estimate the tumor microenvironment. X2/Fisher-Exact were used and significance was determined as P-value adjusted for multiple comparison ( Q < 0.05). Real-world overall survival (rwOS) was obtained from insurance claims and calculated from tissue collection to last contact; time-on-treatment (TOT) was calculated from start to finish of ICI treatments. Results: Overall, 17.7% of HCC expressed PD-L1 by IHC; 79/1306 (6.1%) had high expression. The overall prevalence of dMMR/MSI-H was 0.2%; TMB was high in 5.1%. PD-L1 expression was not associated with MSI-H or high TMB. When comparing tumors that are PD-L1 high vs. negative vs. low, expression of several immuno-oncologic (IO) markers LAG3 (median TPM: 2.4, 0.8, 0.5), CTLA-4 (2.9, 1.2, 0.5), IDO1 (4.2, 1.3, 0.6) and others, as well as T-cell inflamed and IFNg scores all decreased significantly with PD-L1 (all q

287 Background: Addition of immune checkpoint blockade to anti-HER2 therapy has improved outcomes in HER2-positive GEC. Anti-HER2 antibody-drug conjugates have shown activity in some HER2-L tumors in other tumor types. We aimed to compare the molecular profile and TIME of HER2-L and HER-H GEC. Methods: 8678 GEC (gastric, GE junction, and esophageal) adenocarcinoma and squamous cell carcinoma samples were analyzed by next-generation sequencing (NGS) of RNA (whole transcriptome, NovaSeq), DNA (592 genes, NextSeq, or whole exome sequencing, NovaSeq), and immunohistochemistry (IHC, Caris Life Sciences, Phoenix, AZ). Cohorts were stratified by IHC HER2 values of 0 (non-expressors), 1-2+ (HER2-L), or 3+ (HER2-H) and compared using X2 or Fisher-Exact. Statistical significance was determined as P-value adjusted for multiple comparisons (q < 0.05). Microenvironment cell population (MCP) counter was used to quantify immune cell infiltration. Results: Tumor were grouped into HER2 non-expressors (N = 5217), L (N = 2660), and H (N = 801). Mutations of TP53 (72% vs 92%) and amplification of MYC (4% vs 7%), CCNE1 (7% vs 12%), CCND3 (2% vs 4%), CDK6 (3% vs 6%), SMARCE1 (2% vs 25%) and RARA (3% vs 24%) were significantly lower in HER2-L compared to HER2-H (q < 0.05). ARID1A (14% vs 9%), PIK3CA (8% vs 3%), KRAS (9% vs 2%), GNAS (2% vs 0.3%), KMT2D (6% vs 1%), CDH1 (5% vs 1%), and ATM (4% vs 1%) mutations were significantly higher in HER2-L compared to HER2-H (q < 0.05). HER2-L was associated with more TMB-H (9.3% vs. 5%; q 0.05). Immuno-oncology (IO)-related gene expression inversely correlated with HER2 expression with lowest expression of PDCD1LG2, CD274, CTLA4, PDCD1, HAVCR2, CD80, IFNG, LAG3, and CD86 in HER2-H (q < 0.05). HER2-L had significantly higher median immune infiltration of B cells (fold change [FC]: 1.22), T cells (FC: 1.16), CD8+ T cells (FC: 1.56), NK cells (FC: 1.12), neutrophils (FC: 1.10), cytotoxic lymphocytes (FC: 1.36), and myeloid dendritic cells (FC: 1.34), compared to HER2-H (q < 0.05). HER2 non-expressors showed similar immune cell infiltrates compared to HER2-H. HER2-H was associated with lower T-cell inflamed scores and IFN gamma signature when compared to HER2-L and non-expressors (q < 0.05). Conclusions: To our knowledge, this study is the first and largest comparison of molecular profile and TIME in HER2-expressing GEC. We demonstrated distinct molecular and TIME profiles with higher immunogenic profiles in HER-L as compared to HER2-H. IO-related gene expression and TIME cell distribution differences in HER2-H GEC suggest that response to IO and HER2 therapy combinations is likely related to HER2-targeted treatment effect on TIME rather than baseline immunogenicity of the tumor.

Human control of multiple agents in swarm or multi-agent deployments are still novel and understudied—especially when real-world research involving many agents (e.g., 50+) are considered, though many organizations are rushing to deploy them for diverse purposes from package delivery to military engagements. Challenges include the scale of situation awareness, workload and attention splitting associated with division of labor across many functions, human perceptions of emergent swarm behaviors, etc. The challenges in either fielding many automated agents in real world settings using current technologies, or of accurately simulating those agents and behaviors sufficiently are posing difficulties—and will become ever more difficult as scale and complexity increases. This panel will focus on recent efforts to create and evaluate methods to support human control of large numbers of swarm or multi-agent vehicles, with an emphasis on deployment and lessons from real world “in the wild” evaluations using multiple representative machine agents.

In this paper, we study human trust and its computational models in supervisory control of swarm robots with varied levels of autonomy (LOA) in a target foraging task. We implement three LOAs: manual, mixed-initiative (MI), and fully autonomous LOA. While the swarm in the MI LOA is controlled by a human operator and an autonomous search algorithm collaboratively, the swarms in the manual and autonomous LOAs are fully directed by the human and the search algorithm, respectively. From user studies, we find that humans tend to make their decisions based on physical characteristics of the swarm rather than its performance since the task performance of swarms is not clearly perceivable by humans. Based on the analysis, we formulate trust as a Markov decision process whose state space includes the factors affecting trust. We develop variations of the trust model for different LOAs. We employ an inverse reinforcement learning algorithm to learn behaviors of the operator from demonstrations where the learned behaviors are used to predict human trust. Compared to an existing model, our models reduce the prediction error by at most 39.6%, 36.5%, and 28.8% in the manual, MI, and auto-LOA, respectively.

Over the last decade interest has been shown in people with symptomatic lung disease who have features both of COPD and asthma. In this review we examine how COPD and asthma are defined and examine clinical characteristics of people defined by researchers as having asthma-COPD overlap (ACO). We look at pathological and physiological features along with symptoms and consider the impact of each diagnosis upon therapeutic management. We highlight challenges in the diagnosis and management of airway disease and the various phenotypes that could be part of ACO, in so doing suggesting ways for the clinician to manage patients with features of both asthma and COPD.

Background: Acral lentiginous melanoma (ALM) is a rare melanoma subtype found on the palms, soles and nailbeds. Outcomes are poor for patients with advanced ALM, and novel treatment approaches are needed. Here, we seek to explore the global genomic and transcriptomic landscape of ALM. Methods: A total of 699 primary CM (non-ALM cutaneous melanoma) and 18 primary ALM samples underwent next generation sequencing of DNA (592 Gene Panel, NextSeq, or WES, NovaSeq), and whole transcriptome sequencing (NovaSeq, WTS). Wilcoxon, Fisher’s exact test were used to determine statistical significance (displayed as p value without and q value with multi comparison correction). xCell, HLA subtyping, neoantigen load (HBA: high binding affinity; IBA: intermediate binding affinity; LBA: low binding affinity), Interferon gamma score (IFNγ), MAPK pathway activity score (MPAS), and Innate anti-PD-1 Resistance score (IPRES) were calculated by mRNA expression. Global differentially regulated genes were assessed via limma R package (C: log fold change). Results: The most common alterations in ALM included NRAS (22.2%), NF1 (20.0%), BRAF (11.1%) and CDKN2A (11.1%) mutations, and EMSY (22.2%), ELL (11.1%), MAML2 (11.1%), MRE11(11.1%) and PIK3R2 (11.1%) amplifications. ALM had lower TMB (1.5 v 9 Mut/Mb, q Conclusion: ALM has distinct immunologic features, including upregulation of HLA-G, as well as lower MAPK activation in ALM, compared to CM, highlighting the need for novel therapeutic approaches in the treatment of this rare subtype. Citation Format: Gino K. In, Jun Yin, Phillip Walker, Justin Moser, Joanne Xiu, Kelsey Poorman, Geoffery T. Gibney, Matthew Oberley, Thuy Phung, Leonel F. Hernandez-Aya, Jose Lutzky, Wolfgang Michael Korn, Michael B. Atkins. Comprehensive genomic and transcriptomic profiling of acral lentiginous melanoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6130.

Background: Activation in RAS pathway has been associated with cancer development. Three RAS family members, including NRAS, KRAS and HRAS are frequently mutated across various cancer types, where NRAS mutations are present in 15-20% of melanomas. NRAS-mutant melanomas (NRASm) have been extensively characterized. However, molecular and clinical implications of HRAS mutations (HRASm) in melanoma are less well understood. Methods: A total of 6329 melanoma samples were subjected to comprehensive molecular profiling at Caris Life Sciences. Analyses included next generation sequencing of DNA (592 Gene Panel, NextSeq; whole exome sequencing, NovaSEQ), RNA (NovaSeq, whole transcriptome sequencing, WTS) and IHC. MPAS scores to evaluate MAPK pathway activation, IFN scores, QuantiSeq, neoantigen load (high, intermediate, low binding affinity: HBA, IBA and LBA) and GSEA were calculated from mRNA expression data. Wilcoxon, Fisher’s exact were used to determined statistical significance (p value without and q value with multi comparison correction; FDR for GSEA). The reference cohort was the entire melanoma cohort (MC). Results: HRASm were identified in 69 (1.09%) of melanoma samples (hotspots mutations: G13, 40%; Q61, 34%; G12, 18% and others, 9%). HRASm and NRASm had different genomic landscapes: HRASm were significantly associated with a higher mutation rate of NF1 (43.2% vs 27.7%, p Conclusions: The genomic landscape of HRASm are significantly different from that of NRASm, implying their distinct roles in tumorigenesis. HRASm also demonstrated higher MAPK activation, suggesting that they could potentially benefit from agents targeting on this pathway. In addition, HRASm displayed more immunogenic features, associated with down-regulation of angiogenesis pathway, revealing a potential higher susceptibility of HRASm to immunotherapy. Citation Format: Leonel F. Hernandez-Aya, Estelamari Rodriguez, Aparna Nallagangula, Jun Yin, Phillip Walker, Joanne Xiu, Justin Moser, Gino K. In, David Spetzler, Geoffery T. Gibney, Matthew Oberley, Thuy Phung, Michael Atkins, Dave S. Hoon, Wolfgang Michael Korn, Jose Lutzky, Gilberto Lopes. Molecular and immunologic characterization of HRAS mutations in a cohort of 6,329 patients with cutaneous melanoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5625.

11551 Background: Immune checkpoint inhibitors (ICI) have limited efficacy for most sarcomas. Yet, responses are seen in particular sarcoma subtypes, highlighting the need for better predictive biomarkers. The T cell inflamed score (TIS), a gene expression signature reflective of an active tumor immune microenvironment, is associated with ICI response in multiple solid tumors. We evaluated the TIS across a large database of sarcomas to identify which histologic subtypes may benefit from ICI. Methods: Next generation sequencing of DNA (592 gene or whole exome)/RNA (whole transcriptome) was performed for 3605 sarcoma patient samples, representing 45 histologic subtypes (Caris Life Sciences, Phoenix, AZ). TIS (18 gene weighted coefficient composite value; Cristescu 2018) was calculated and the Microenvironment Cell Populations-counter tool (Becht 2016) was used to quantify immune cell populations. Results were compared to melanoma (n = 1255), a representative immunogenic tumor type. High TIS was defined as a score within the upper quartile of melanoma TIS (> 5.5). Percentage with high TIS are reported with 95% CI. Results: Median TIS was highest in inflammatory myofibroblastic tumor (IMT), epithelioid sarcoma (EPIS), myxofibrosarcoma (MFS), well differentiated liposarcoma, and solitary fibrous tumor (SFT). These did not differ significantly from melanoma (p > 0.06). Median TIS was lowest in embryonal rhabdomyosarcoma, desmoid tumor (DES), synovial sarcoma (SYNS), and Ewing sarcoma (ES). Histologic subtypes where > 10% of samples had a high TIS included IMT (29.9% ± 21.7%), MFS (23.3% ± 12.6%), pleomorphic sarcoma (PLSARC) (21.9% ± 5.8%), cutaneous angiosarcoma (ANGS) (18.4% ± 13.9%), spindle cell sarcoma (17.5% ± 7.6%), liposarcoma (LPS) (17% ± 10.7%), EPIS (15.4% ± 19.6%), visceral ANGS (13.2% ± 10.7%), pleomorphic LPS (13.6% ± 14.3%), fibrosarcoma (12.5% ± 13.2%), leiomyosarcoma (11.6%± 3.4%), malignant peripheral nerve sheath tumor (MPNST) (10.2% ± 7.7%), and perivascular epithelioid cell tumor (PEComa) (10% ± 10.7%). The relative abundance of immune and stromal cell populations was highly variable across sarcoma subtypes, yet a strong positive correlation between TIS and immune cell populations was observed for most subtypes (e.g. T cells, Spearman R range: 0.56 [P = 0.08] - 0.96 [P < 0.0001]). A notable exception was SFT, which had a relatively high median TIS but low abundance of CD8+ T cells and B cells. Conclusions: We found high median TIS and/or significant proportions of samples with a high TIS in sarcoma subtypes with previously demonstrated responsiveness to ICI, including MFS, PLSARC, LPS, and ANGS, while unresponsive tumor types such as RMS, DES, SYNS, and ES had low TIS. We further identified subtypes with high TIS but limited prior clinical data supporting ICI use, such as IMT, EPIS, MPNST, SFT, and PEComa. Our results warrant prospective exploration of TIS as a predictive biomarker for ICI use in sarcoma.

561 Background: The RAS pathway regulates tumorigenesis and cell proliferation. HRAS is a RAS family member that activates via farnesylation. Indirectly targeting mutant HRAS with tipifarnib, a farnesyltransferase inhibitor (FTI), recently demonstrated efficacy in head and neck tumors. We aimed to investigate the molecular characteristics and clinical outcomes of HRAS mutations (HRASmut) for any potential role as a prognostic and therapeutic biomarker in breast cancer (BC). Methods: A total of 14,013 BC tissue samples had molecular profiling, including next generation DNA (592 Gene Panel, NextSeq, or WES, NovaSeq) or RNA sequencing (NovaSeq, WTS), and immunohistochemistry analyses, at Caris Life Sciences. MAP kinase (MAPK) activation and likelihood of a tumor’s response to anti-PD1 therapy were evaluated via MAPK Pathway Activity Score (MAPS) and interferon (IFN) score, respectively. Wilcoxon, Fisher’s exact, or Dunnett’s tests were used to determine statistical significance. Overall survival (OS) was calculated from date of tissue collection to insurance claims last contact using the Kaplan-Meier method. HRAS mutations (HRASmut) were compared to the general BC cohort (GC). Results: HRASmut were significantly enriched in older patients (median 69 vs 60 yrs; q

1041 Background: Amplification or overexpression of human epidermal growth factor receptor 2 (HER2) oncogene is present in about 15-20% of breast cancers & is a prognostic & predictive biomarker. Additional ERBB2/HER2 alterations have become apparent on tumor next generation sequencing (NGS), including activating kinase domain mutations & fusions. Methods: DNA NGS (592 gene panel or whole exome) data from 12,153 breast samples retrospectively reviewed for ERBB2 alterations with RNA whole-transcriptome sequencing (WTS) data available for 7289 (60%) samples. Gene fusions detected using the ArcherDx fusion assay or WTS. Clinicopathologic features were described including breast cancer subtype, age, & biopsy site. HER2 status determined according to 2018 ASCO-CAP guideline. Overall survival obtained from insurance claims & Kaplan-Meier estimates were calculated for defined patient (pt) cohorts. Statistical significance was determined using Chi-square & Wilcoxon rank sum tests. Results: ERBB2 mutations ( ERBB2mts) were identified in 3.2% (n = 388) of tumors overall & most common in liver metastases (113/1972, 5.7%). ERBB2mts were found more in breast lobular tumors compared to ductal tumors (10 vs 2.1%, p < 0.001). HER2+ tumors had higher frequency of ERBB2mts compared to HER2- (4.3 vs 3%, p = 0.028). Tumors with score of 0 by immunohistochemistry demonstrated lower rate of ERBB2mts (0+ 2.2%, 1+ 3.5%, 2+ 4.5%, 3+ 3.45%, p < 0.05). Among HER2- tumors, ERBB2mts were present in 3.6% of hormone receptor (HR)+/HER2- & 1.9% of TNBC. Metastatic tumors had a higher rate of ERBB2mts compared to locoregional breast tumors (3.8 vs 2%, p < 0.001), with increased rates of activating mutations S310F (0.1 vs 0.0%, p < 0.05) & D769H (0.3 vs 0.1%, p < 0.05), & the resistance mutation L755S (1.2 vs 0.6%, p < 0.01). Compared to ERBB2-WT, ERRB2mts were associated with decreased ERBB2 transcripts levels in HER2+ samples (222 vs 441 transcripts per million [TPM], p < 0.001) & increased levels in HER2- samples (73 vs 35 TPM, p < 0.001). High tumor mutational burden (≥ 10 mut/Mb) & ERBB3 mutations were more common in ERBB2mts compared to ERRB2-WT (16.7 vs 7.7%, p < 0.001; 10.6 vs 0.8%, p < 0.001). ERBB2 fusions were rare (0.49%) with 97% occurring in HER2+ tumors. Of 8358 pts with outcome data, prognosis (HR 1.2, P = 0.06) & response to chemotherapy (HR 1.1, P = 0.42) was similar between pts with HER2- ERBB2mt & ERBB2-WT. Conclusions: ERBB2mts & fusions were observed in all breast cancer subtypes - more commonly in HER2+, metastatic, & lobular histology tumors - & did not influence prognosis. These alterations may reflect response to treatment pressures in HER2+ disease to reactivate HER2-mediated growth pathways following anti-HER2 therapy & may represent a targetable upregulated oncogenic pathway in HER2- disease. Ongoing identification of ERBB2 alterations may augment treatment options for breast cancer pts & clinical outcomes from this approach are under investigation.

9086 Background: NF1 is a tumor suppressor gene that regulates the RAS-MAPK and mTOR pathways. Co-mutations previously observed with NF1-mutations (mt) include TP53, KRAS, EGFR and rarely HER2, STK11, and PIK3CA mutations. We report a comprehensive molecular characterization with clinical outcomes analyses for NF1-mt non-small cell lung cancer (NSCLC). Methods: Next-generation sequencing (NGS) of DNA (592-gene or whole exome) and RNA (whole transcriptome) was performed for NSCLC patient (pt) samples (n = 10,310) submitted to a CLIA-certified laboratory (Caris Life Sciences, Phoenix, AZ).. RAS-MAPK and PI3K-AKT-MTOR signaling were assessed by transcriptional signatures of pathway activation (MAPK pathway Activation Score [MPAS], Wagle, 2018; and GSEA Hallmarks collection, respectively)., PD-L1 by immunohistochemistry (IHC, positive: TPS ≥1%), high tumor mutational burden (TMB) defined as ≥10 mut/Mb, and deficient mismatch repair/high microsatellite instability (dMMR/MSI-High) was assessed by IHC/NGS. Overall survival (OS) was obtained from insurance claims. Statistical significance was determined using Chi-square & Wilcoxon rank sum tests. P-value adjust for multiple hypothesis testing (Benjamini-Hochberg). Results: NF1-mt were identified in 1,045 NSCLC samples (10.1%). Concurrent KRAS, EGFR, ERBB2, BRAF or MET alterations are noted in Table 1, with no ROS1, RET or ALK fusions identified. Compared to NF1-wt, NF1-mt NSCLC was associated with increased RAS-MAPK expression (3.0-fold, P < 0.0001), while PI3K-AKT-MTOR-signaling was not significantly increased (2.1-fold, P = 0.12). Rates of TMB-High (51.7% vs 32.5%, P < 0.0001), PD-L1+ (69.1% vs 58.8%, P = 0.06), and dMMR/MSI-High (1.7 vs 0.7%, P < 0.05) were higher in NF1-mt samples. OS and duration on treatment from the start of Pembrolizumab (HR: 1.0 and 1.0, respectively) or other IOs (HR: 0.9 and 1.0, respectively) were not significantly different between NF1-mt and NF1-wt patients, However, among NF1-mt samples, high TMB and TP53-wt were associated with better OS (HR 0.6, P < 0.05 each). Conclusions: NF1-mt patients rarely harbored actionable NSCLC driver co-alterations. NF1-mt cases showed increased activation of RAS-MAPK axis, which may represent a potential pathway to target with MEK inhibitors. NF1- mt are responsive to immunotherapy and better outcomes are seen with high TMB and absence of TP53 mutations. Further work is warranted to determine the influence of actionable drivers on targeted therapy outcomes in NF1-mt NSCLC. [Table: see text]

9119 Background: MET exon 14 skipping mutations (METex14) join a growing list of viable therapeutic targets in advanced NSCLC. Several unique features distinguish METex14 from other established targets. METex14 has been characterized as a tumor-agnostic genomic alteration, though most frequently reported in lung adenocarcinoma. However, METex14 represents a family of mutations (mt), not a single alteration, and there is notable heterogeneity in histology. The degree and significance of heterogeneity within METex14 have not been well characterized. Methods: NSCLC tissue samples were analyzed with DNA-based next-generation sequencing (NGS; 592 genes, NextSeq) or whole-exome sequencing (NovaSeq), RNA-based whole transcriptome sequencing (WTS, NovaSeq), and immunohistochemistry (IHC) at Caris Life Sciences (Phoenix, AZ). PD-L1 expression utilized the 22C3 clone (Dako); TMB-high was defined as ≥ 10 mt/Mb. Wilcoxon or Fisher’s exact were used to determine statistical significance (p without and q with multi comparison correction). Immune cell fraction (QuanTiseq) and pathway analysis (ssGSEA) were informed by WTS analysis. Results: A total of 440 METex14 cases were identified: 49 (11.1%) with squamous histology, 381 (86.6%) with non-squamous histology, and 10 (0.2%) with adenosquamous histology. A total of 147 distinct METex14 mutations were detected. The most common METex14 mutations were D1028H (8.4%), D1028N (7.0%), c.3082+2T > C (5.7%), D1028Y (5.2%), and c.3082+1G > A (4.5%). Co-mutations in TP53 were common (43.9%) but varied by specific METex14 mutation; TP53 co-mutations were observed in 53.9% of c.3082+3A > T but only 21.1% of c.3082+1G > T. Among all METex14 cases, 8.6% were TMB-high, but this varied by specific mutation with a median TMB of 2 mt/Mb in MET c.3082+2T > A and a median of 7 mt/Mb in MET c.3082+1G > C (q < 0.05). PD-L1 expression ≥ 1% was present in 82.2% of METex14 samples but also varied by specific METex14 mutation with a median PD-L1 tumor proportion score (TPS) of 97.5% in MET c.3082+1G > C and a median TPS of 0% in MET c.3082+3A > G (q < 0.05). Co-mutations varied by histology: in squamous METex14, 90.4% had TP53 mt (p < 0.001), 17.9% had KMT2D mt (p < 0.05), and 10.7% had PIK3CA mt (p < 0.05), while in non-squamous METex14, 60.7% had TP53 mt, 2.7% had KMT2D mt, and 4.3% had PIK3CA mt. Survival was numerically shorter in squamous METex14 NSCLC compared to non-squamous (HR 1.22, p = 0.47, mOS 336 vs.1106 days). Conclusions: There is significant heterogeneity within METex14 NSCLC with differences in co-mutations, TMB, and PD-L1 expression noted among different METex14 mutations. While METex14 is detected in both squamous and non-squamous NSCLC, there are differences in enrichment of oncogenic pathways. The clinical impact of these differences warrants further investigation.

11580 Background: Synovial sarcomas (SS) harbor a specific, balanced, reciprocal translocation t(X;18) leading to the oncogenic SS18-SSX fusion. Defective DNA damage response (DDR) is a hallmark of cancer leading to genomic instability and is associated with chemosensitivity. Efforts have been made to identify a genetic signature that predicts SS progression, treatment response, and survival in order to identify more accessible and effective treatments. This investigation explores the role of DDR in pathogenesis of SS. Methods: Patients with the diagnosis of SS from 2013 to 2021 within the Caris Life Science database were included in the study. A combination of NGS of DNA and RNA at a CLIA-certified laboratory (Caris Life Sciences, Phoenix, AZ) was performed on archival tumors. Homologous recombination deficiency (HRD) scores were calculated as a composite of loss of heterozygosity, telomeric allelic imbalance, and large-scale transitions, using a positive threshold of 42. This study was approved by the University of Miami (UM)-Sylvester IRB and all the data collected was de-identified. Results: A total of 120 patients were identified with 49 of these patients from UM-Sylvester. Mean age of diagnosis on the sample was 46 years old (range of 15-86) and 45.8% were female. Among the 49 patients from UM-Sylvester, mean age was 60 years old (range of 35-84), 28 patients had a gene alteration identified (57%) and 6 of them a homologous recombination deficiency (HRD) gene (12%). A total of 63 different genes were identified with the most common TP53 (49%), LOH (12.2%), ATRX (10.2%) and RB1 (6.1%). Other HRD genes identified were MLH1 (4%) and CHEK2 (4%). There was no correlation identified between the age (15-65 vs elderly ⩾ 65 years) or the gender (female vs male) and the presence of a mutated DDR (p = 0.615; p = 0.091 respectively). Within the entire Caris database (N = 120), we identified 11 patients (N = 120, 9%) whose tumor tested positive for any DDR gene alteration. The most common were ATM (2.6%), followed by ATRX (1.6%) and CHEK2 (0.9%). The median HRD score was 22 within the sample. Conclusions: We report here the most common genes altered on molecular profile for a large cohort of SS samples. The prevalence of predicted pathogenic DDR gene mutation carriers in our cohort (9%) suggests that constitutional defects in this pathway may be associated with SS. We found higher rates of positive DDR on the UM-Sylvester cohort, and this could be associated with specifics of our population, given high frequency of Hispanic patients. Work is ongoing to associate our findings with race, ethnicity, survival and response to treatments. Work is ongoing to associate our findings with race, ethnicity, survival and response to treatments. These correlations will be reported in the final abstract. Cytotoxic therapy remains gold standard for metastatic SS, but better understanding of the molecular profile can pave way for further options, including targeted therapy and immunotherapy.

5548 Background: Clear cell carcinomas (CCC) are rare histologies outside of the kidney and are typically less sensitive to standard treatments. Genomic alterations in chromatin remodeling pathways involving ARID1A or the intracellular PI3K-mTOR signaling pathway are found in both renal and ovarian CCC. It is unclear whether CCCs originating from different anatomic sites share a common genomic landscape. This CARIS Precision Oncology Alliance project sought to determine whether CCC of different organs shared similar genomic signatures and to identify potential pathways that could be targeted in a tumor-agnostic clinical trial. Methods: CCCs (N = 861) from multiple primary tumor sites, including kidney (30.5%), ovary (39%), endometrium (23.9%), other gynecologic sites (e.g., cervix, fallopian tube, 3.3%), and miscellaneous (non-kidney or gynecologic sites, 3.3%) were analyzed at the Caris Life Sciences Laboratory (Phoenix, AZ). Using hierarchical clustering (HC) and principal component analysis (PCA), the samples were compared across 648 total genes from five metabolic gene sets consisting of angiogenesis, glycolysis, hypoxia, oxidative phosphorylation, and fatty acid metabolism. Gene Set Enrichment Analysis (GSEA) was further conducted on the samples across fifty hallmark gene sets representing specific biologic processes and expression. Samples were also analyzed for individual genomic alterations and immune-oncology associated biomarkers. PD-L1 (SP142) expression was evaluated by immunohistochemistry (positive threshold: 2+ stain intensity and ≥ 5% tumor cells). Results: HC and PCA demonstrated that renal CCC formed distinct clusters compared to non-renal CCC. Tumors from gynecologic sites could not be separated into distinct clusters. GSEA showed that the hypoxia gene set was significantly upregulated in the renal but not in non-renal CCCs. Mutations involving TP53, ARID1A, PIK3CA were found to be the most altered genes in endometrial (62%, 26%, 31%), ovarian (13%, 55%, 48%), other gynecological sites (33%, 38%, 44%), and non-gynecologic CCC (13%, 17%, 12%) respectively. PD-L1 expression, high tumor mutational burden (≥10 mutations/Mb), and deficient mismatch repair/microsatellite instability rates across sites were: kidney (11%, 2%, 2%), endometrium (13%, 12%, 7%), ovary (9%, 4%, 3%), other gynecological sites (31%, 11%, 11%), and miscellaneous sites (11%, 19%, 4%). Conclusions: Initial metabolic gene expression clustering analysis shows that CCCs do not separate by organ of origin beyond renal versus extra-renal. TP53, ARID1A, and PIK3CA were the most frequently altered genes in non-renal CCC. Out of fifty hallmark gene sets, only two were statistically significantly different among gynecological CCCs. This similarity between gynecological CCC can be leveraged by targeting pathways such as PI3K-AKT-mTOR, DNA repair, and MYC targets in a site agnostic manner. Furthermore, high PD-L1 expression is found in other gynecological sites.

2629 Background: High YAP1 expression correlates with the ‘T-cell inflamed’ expression phenotype in small cell lung cancer (SCLC), but its association with other biomarkers of immune checkpoint vulnerability and in tumor types beyond SCLC is not known. We examined whether YAP1 expression correlates with other established markers of immune checkpoint blockade (ICB) efficacy (PDL1 expression and TMB) in a tumor agnostic manner to determine clinical relevance. Methods: Next-generation sequencing of DNA (592 gene panel or whole exome) and RNA (whole transcriptome) was performed for patient samples (n = 57,134), representing 13 cancer types, submitted to a CLIA-certified laboratory (Caris Life Sciences, Phoenix, AZ). The ‘T-cell inflamed’ signature (TIS) score was calculated as an 18-gene weighted coefficient composite value (Cristescu, 2018). PDL1 expression was assessed by immunohistochemistry (IHC) with cancer type-specific antibodies and thresholds, and high tumor mutational burden was defined as ≥10 mut/Mb. Patients were stratified into subgroups based on median YAP1 expression (YAP1-High/YAP-Low) within each cancer type. Significance was tested by Chi-square, Fisher’s exact test, or Mann-Whitney U test. Results: YAP1-High tumors were associated with significantly increased TIS scores compared to YAP1-Low across all 13 cancer types examined, with the largest fold increase observed in SCLC (1.33-fold, p < 0.0001), followed by pancreatic cancer (1.28-fold, p < 0.0001), while the smallest occurred in melanoma (1.13-fold, p < 0.0001). Spearman correlation strength (range 0.23-0.57) between YAP expression and TIS scores was consistent with increased TIS scores in YAP1-High samples. TMB-High rates were similar in YAP-High and YAP1-Low subgroups for most cancer types, with slightly lower rates in YAP1-High tumors observed for endometrial (23.0 vs 26.6%, p < 0.001) and esophageal (7.0 vs 9.5%, p < 0.05) cancers. YAP1 expression was not significantly increased in PDL1+ (IHC) tumors for most cancer types. However, significantly decreased YAP1 expression was associated with PDL1+ samples in RCC (Renal Cell Carcinoma) (0.91-fold change, P < 0.05), MM (0.90-fold change, P < 0.001), and ENCA (0.80-fold change, P < 0.0001) Conclusions: Our analyses provide confirmation that YAP1 expression positively correlates with the ‘T-cell inflamed’ phenotype across many cancer types, including those with approvals for (ICB) therapy. YAP1 expression was independent of established markers of ICB response, including TMB and PDL1. Further analysis of YAP1 expression as an additional tumor agnostic predictive biomarker is warranted.

9122 Background: Multiple DNA alterations in exon 14 splice sites have been identified in NSCLC and result in skipping of the juxtamembrane domain Cbl-E3 ubiquitin ligase binding region, leading to increased MET stability and oncogenesis. The effects of these alterations on transcriptome-level have not been fully characterized. We present the largest cohort study of METex14 using WTS and identify key cellular pathways associated with invasion and metastases in METex14. Methods: 17,666 NSCLC tumor samples underwent genomic profiling at Caris Life Sciences. Analyses included next generation sequencing of DNA (592 Gene Panel, NextSeq, whole exome sequencing, NovaSeq) and RNA (NovaSeq, WTS). METex14 was captured via WTS. ssGSEA analysis was used to evaluate pathway enrichment. Wilcoxon, Fisher’s exact were used for statistical significance (p without and q values with multiple comparison correction). Results: 440 patients (2.5%) with METex14 were identified. METex14 patients were of older age, female gender, and enriched in sarcomatoid histology (Table 1). The most common alterations were point mutations (51.5%) and deletions (17.3%) at donor splice sites. Splice site alterations except point mutations at splice acceptor site translated to increased mRNA expression compared to wild-type MET (WT). MET amplification translated to higher mRNA expression compared to METex14 and WT with synergistic expression when co-altered with METex14 (q 50% (53% vs. 27.6% WT, q

4142 Background: The composition of the tumor microenvironment (TME) in PDACs is more heavily driven by mutant (mt) KRAS than any other cancer. How genomic alterations of PDAC including KRAS status affect the immune cell (IC) landscape remains unclear. Thus, we characterized IC types and the prevalence of immuno-oncologic (IO) biomarkers in PDAC by genomic and transcriptomic analysis, and investigated associations of mt KRAS with IC estimates in the TME. Our findings were compared to our previous study in CRC. Methods: A total of 4,142 PDAC and 3,727 CRC with KRAS- mts were analyzed using next-generation DNA sequencing (NextSeq, 592 gene panel or NovaSeq, WES), IHC, and whole transcriptome RNA sequencing (NovaSeq) (Caris Life Sciences, Phoenix, AZ). MSI/MMR was tested by FA, IHC and NGS. TMB-H was classified based on a cut-off of >10 mutations per MB. ICs were estimated by QuantiSeq (Finotello 2019, Genome Medicine) or MCP counter (Betcht 2016, Genome Biology). Significance was determined by X2 and Fisher-Exact and p-adjusted for multiple comparisons (q+ & CD8+ T cells, T-reg, NK, myeloid dendritic and endothelial cells compared to KRAS wt. In CRC, a similar pattern was observed but more pronounced in PDAC. Immune-regulatory markers, were among multiple genes downregulated in KRAS-mt PDAC, including CTLA-4 and LAG3. Overall changes were most pronounced in cases harboring KRAS G12D, G12V, Q61, and rare KRAS variants. Conclusions: The TME of KRAS mt PDAC shows IC patterns similar to KRAS mt CRC. Actionable IO-targets, such as PDL1, are enriched in tumors harboring specific variants of KRAS mt PDAC including the targetable G12C variant. If G12D becomes druggable, it could be targetable in 35% patients with PDAC or 15% in CRC. These results demonstrate that the TME of PDAC and CRC shows immune-cold features. Tailored immunotherapeutic strategies would have to overcome these barriers in KRAS mt PDAC and CRC, possibly in combination with molecularly targeted treatment strategies.

2633 Background: Older patients (pts) with cancer are underrepresented in registrational clinical trials for immune checkpoint inhibitor (ICI) therapies. There may be relevant differences in the makeup of the tumor microenvironment (TME) and in genomic signatures of cancer in older pts. This analysis explores differences in the genomic makeup of common cancers and their TME in pts ≥ 80 years (yr) of age, compared to younger pts. Methods: Next-generation sequencing of DNA (592 gene panel, NextSeq or whole-exome sequencing, NovaSeq) and RNA (whole transcriptome sequencing, NovaSeq) was performed for non-small cell lung carcinoma (NSCLC; n = 19,891), melanoma (MEL n = 2,899), and renal cell carcinoma (RCC; n = 1,333) pt samples submitted to a CLIA-certified laboratory (Caris Life Sciences, Phoenix, AZ). PD-L1 expression was assessed by immunohistochemistry (IHC), and high tumor mutational burden (TMB-H) was defined as ≥10 mut/Mb. Pts were stratified into age subgroups of ≥80 and < 80 yr for comparison of DNA damage response (DDR) gene alterations, gene expression profiling, and TME analysis (MCP-counter; Becht, 2016). P-values were adjusted for multiple hypotheses testing (Benjamini-Hochberg) unless noted as exploratory. Results: Pts ≥80 yr accounted for 16.0%, 19.9% and 5.3% of NSCLC, MEL and RCC pts, respectively. Compared to pts < 80 yr, NSCLC and MEL pts ≥80 yr had similar DDR gene mutation rates, while BRCA1 mutations were more common in MEL pts ≥80 yr (2.1 vs 0.8%; exploratory- p < 0.05). NSCLC ≥80 yr TMEs had increased abundance of fibroblasts (1.09-fold, p < 0.01), dendritic cells (1.07-fold, p < 0.01) and macrophages (1.04-fold, p < 0.01), and MEL≥80 yr TMEs had fewer infiltrating T-lymphocytes (0.87-fold, p = 0.02). Increased expression of immune checkpoint (IC) genes PDCDL1G2 (PD-L2; 1.11-fold), HAVCR2 (TIM-3; 1.11-fold ), and CD80/86 (1.07/1.08-fold, p < 0.05) was seen in NSCLC pts ≥80 yr, while IL-6 expression was decreased (0.88-fold; p < 0.05). The largest change in IC gene expression was for IL-6 (1.24-fold, p = 0.78) in MEL, and GZMB (0.56-fold ; p = 0.17) in RCC ≥80 yr. TMB-H was less common in NSCLC (29.7 vs 36.5%, p < 0.001) and more common in MEL pts ≥80 yr (65.7 vs 49.0%, p < 0.01), and PD-L1 (IHC-SP142, ≥2+|5%) expression was less frequent in RCC pts ≥80 yr (9.1 vs 19.4%, exploratory p < 0.05). Profiling of glutamine and glucose metabolism-related genes revealed increased SLC38A5 (1.17-fold; p < 0.0001) and decreased G6PC (0.65-fold, p < 0.01) expression in NSCLC ≥80 yr. While not statistically significant, MEL and RCC pts ≥80 yr had opposite trends for SLC38A5 and G6PC expression. Conclusions: Our analysis provides new insights to immune landscape of NSCLC, MEL, and RCC pts ≥80 yr. Differential gene expression and TME composition changes in this population suggest unique, cancer-specific therapeutic opportunities, and a potential to explore biomarkers of response to ICIs.

11548 Background: Alterations in DNA damage response (DDR) pathways contribute to genomic instability and malignant progression and have been shown to be of clinical significance in several carcinomas and solid tumors. While some studies have identified ostensibly pathogenic variations in known and novel cancer genes with implications for sarcoma risk and treatment opportunities, there is limited information regarding the role of DDR pathway alterations in sarcoma. We identified a gene alteration in ERCC2, a gene that codes for a DNA helicase in the nucleotide excision repair pathway, in a patient with multiply relapsed epithelioid sarcoma (ES), prompting an investigation of DDR pathway alterations in sarcoma samples using a global next-generation sequencing (NGS) platform. Methods: Sarcoma patient samples (N = 5310), representing 38 pediatric and adult histologic subtypes, underwent NGS of DNA (592 gene panel or whole exome) and RNA (whole transcriptome sequencing, N = 3612) at a CLIA-certified laboratory (Caris Life Sciences, Phoenix, AZ). A threshold of 10 mut/Mb was used to identify high tumor mutational burden (TMB-H). IHC was performed for PD-L1 (SP142; 2+|5% = positive). Homologous recombination deficiency (HRD) scores were calculated as a composite of loss of heterozygosity, telomeric allelic imbalance, and large-scale transitions, using a positive threshold of 42 (N = 2138). HRD score association with biomarker status was evaluated overall and in sarcoma subtypes. Results: A pathogenic DDR pathway mutation was noted in 842 (15.9%) of the total samples. ATRX was by far the most commonly altered DDR gene (10% of all samples), with mutations observed across 25 sarcoma subtypes (11 subtypes with > 10% mutation rate: leiomyosarcoma [LMS], perivascular epithelioid cell tumor [PEComa], pleomorphic sarcoma [PLSARC], uterine sarcoma [OUSARC], osteosarcoma, spindle cell sarcoma, angiosarcoma, mesenchymal chondrosarcoma, sarcoma NOS, fibrosarcoma and ES). CHEK2, ATM, and MUTYH mutations were observed in 1-2% of sarcoma samples. More than 20 histologic subtypes showed distinct gene signatures with mutations occurring in > 3% of the samples investigated. ERCC2 was mutated in 3% of ES and 6.5% in PEComa. Median HRD scores ranged between 20-58 across sarcoma subtypes. High rates of deficient HRD (HRDd ≥ 42) were observed in PLSARC (83.2%), OUSARC (73.7%), and dedifferentiated chondrosarcoma (71.4%), while low rates of HRDd were observed in Ewing sarcoma (0%) and clear cell sarcoma (10%). In the overall cohort, ERCC2, ATRX and BRCA2 were significantly associated with increased HRD scores (p = 0.01). Conclusions:DDR pathway alterations are present in numerous histologic subtypes of sarcoma. A more comprehensive analysis of individual histologic subtypes is in progress. Further research will evaluate the clinical implications of these known and novel mutations to guide risk stratification and potential therapeutic options.

11525 Background: Osteosarcoma (OS) incidence is characterized by a bimodal age distribution, with peaks in early adolescence and in adults > 65 years of age. In contrast to adolescents, OS in adults is frequently considered as a secondary neoplasm (i.e., transformation of Paget´s disease of the bone, radiation induced). Yet, the literature is scarce regarding the impact of age on the molecular landscape of OS. Herein, we sought to explore the association between age and the genomic profile as well as the tumor immune microenvironment (TME) in a large cohort of OS patients. Methods: 208 specimens were centrally analysed at the Caris Life Sciences laboratory with DNA seq (NextSeq, 592 gene panel or NovaSeq, whole-exome sequencing), RNA seq (Archer fusion panel or whole-transcriptome sequencing) and immunohistochemistry (IHC). RNA deconvolution and differential expression analyses were performed using the Microenvironment Cell Populations counter method for quantification of immune cell populations and gene expression profiling. The cohort was stratified into three distinct age groups (< 25 years [n = 83], 25-45 years [n = 58], > 45 years [67]). Results: Overall, the most frequently detected mutations were in TP53 (37%), RB1 (13%), ATRX (9%), TERT (6%), PTEN (5%), PIK3CA (4%) and KMT2D (3%). Copy number alterations were most frequently detected in CDK4 (12%), LRIG3 (11%), FLCN (11%), MDM2 (9%), CCND3 (9%), VEGFA (8%), TFEB (8%). Interestingly, age-based stratification revealed an increased frequency of FLCN (19.7 vs 4.7%, p < 0.01), CCND3 (13.9 vs 3.1%, p < 0.05), and HSP90AB1 (11.3 vs 0.0%, p < 0.01), alterations in patients < 25 years compared to > 45 years. TME analysis revealed that patients > 45 years have decreased B-cell abundance compared to patients < 25 years (2.9-fold decrease, p < 0.05) and 25-45 years (4.8-fold decrease, p < 0.05). Although not statistically significant, median transcriptional expression of PD-L1 was numerically increased in patients > 45 years (1.8-fold compared to 25-45 years, p = 0.17; 2.0-fold compared to < 25 years, p = 0.27), which was consistent with increasing rates of IHC PD-L1 expression with age (5.3%, 9.4%, and 17.5%, respectively, p = 0.06). Conclusions: To the best of our knowledge, this study represents the largest cohort of molecularly characterized OS. Age-associated differences in the genetic landscape and TME composition, including increased gene amplifications observed in younger patients and decreased B-cell abundance in older patients, might suggest fundamental underlying molecular and biological differences.

594 Background: Young-onset pancreatic cancer (YOPC; < 50 years at diagnosis) has been associated with male preponderance, extensive smoking history, and a trend towards improved survival compared with average-onset pancreatic cancer (AOPC; ≥70 years). However, the genomic and transcriptomic correlates underlying these clinical differences are incompletely understood. Using a large matched genomic-transcriptomic next-generation sequencing (NGS) dataset, we sought to characterize the distinct molecular landscape associated with YOPC compared with AOPC. Methods: A total of 2430 pancreatic ductal adenocarcinoma NGS samples (YOPC n = 292; AOPC n = 2138) with matched whole-transcriptome (NovaSeq) and DNA (NextSeq, 592-gene or NovaSeq, whole-exome) sequencing data were analyzed (Caris Life Sciences, Phoenix, AZ). Immune deconvolution was performed using the QuantiSeq pipeline. Limited clinical data precluded stage- and treatment-stratified comparisons between cohorts. Overall survival (OS) was obtained from insurance claims, and Kaplan-Meier estimates were calculated for age- and molecularly-defined cohorts. Significance was determined as FDR-corrected P-values (Q) < 0.05. Results: Of 2430 PDAC patients undergoing NGS, YOPC patients (median age 46 years) were more likely to be male (65% vs. 52%; P < 0.001) and current smokers (32% vs. 11%; P = 0.02) compared with AOPC patients (median age 75 years). YOPC patients had higher proportions of mismatch repair-deficient (MMR)/MSI-H (2.8% vs. 0.8%, P = 0.001), BRCA2-mutant (4.7% vs 2.1%, P = 0.009), and PALB2-mutant (1.4% vs 0.5%, P = 0.04) tumors compared with AOPC patients, while tumors in AOPC patients had more frequent SMAD4 (20.1% vs. 14.7%, P = 0.03), RNF43 (6.3% vs. 2.5%, P = 0.012), CDKN2A (24.8% vs. 19.2%, P = 0.04), and SF3B1 (2.7% vs. 0.7%, P = 0.04) mutations. YOPC patients also demonstrated lower HLA-DPA1 homozygosity (55.2% vs. 64.1%, Q < 0.05) vs. AOPC patients. Notably, YOPC patients demonstrated significantly lower incidence of KRAS-mutant (81.3% vs. 90.9%, Q < 0.01) tumors compared with AOPC patients. In the KRAS-wildtype subset (n = 225), YOPC tumors were more likely to be driven by NRG1 and MET fusions, while BRAF fusions were exclusively observed in AOPC patients. Computationally inferred immune deconvolution revealed enrichment of NK cell (Q = 0.04) and M2 macrophages (Q = 0.01) populations in YOPC tumors. There was an association with improved OS in YOPC patients with KRAS-wildtype (median 22.4 [YOPC- KRASWT] vs. 15.1 [AOPC- KRASWT] months, P = 0.02) but not KRAS-mutant (P = 0.28), tumors compared with AOPC patients. Conclusions: In this large real-world multi-omic characterization of age-stratified molecular differences in PDAC, YOPC is associated with a distinct molecular landscape compared with AOPC. These data reveal molecular features of YOPC with prognostic and therapeutic implications.

477 Background: HCC is increasingly prevalent in older adults with rising incidence and an aging population worldwide. Retrospective studies show older patients with HCC may have an increased survival compared to younger patients. However, data is lacking regarding the genomic and biologic differences, that if identified, would potentially change how we treat this disease in younger vs. older patients. Hence, there is a need to better characterize the molecular landscape of the disease in an age-specific manner. We analyzed the association of age with genomic alterations and therapeutic response to sorafenib in a cohort of advanced HCC that had undergone comprehensive molecular profiling. Methods: 487 HCC samples (excluding variants) were analyzed using Next Generation Sequencing (592 gene panel, NextSeq), Whole Exome and Whole Transcriptome Sequencing (NovaSeq), and IHC at Caris Life Sciences (Phoenix, AZ). PD-L1 positivity was determined by IHC (SP-142 clone, cutoff ≥1, 1%). Tumor mutational burden (TMB) was a measure of total somatic mutations per Mb. Immune cell populations were determined by Microenvironment Cell Population (MCP) counter analysis of RNA expression data. Overall survival (OS) calculated from tissue collection to last contact and time on treatment (TOT) with sorafenib were extracted from insurance claims and calculated using Kaplan-Meier curves. Statistical analysis was done using Chi-square, Fisher Exact and Wilcoxon rank sum tests, with p values adjusted for multiple comparisons and q77: A3 (n=75)). With age, mutational frequencies in CTNNB1 (A1=13.04%, A2=33.43%, A3=38.24%) and TERT (A1=25%, A2=68.84%, A3=76.92%) increased, while ATM (A1=6.52%, A2=0.93%, A3=1.49%) decreased (p0.05). There were fold increases in median TMB (A2/A1=1.33, A3/A1=1.33, p65 (p=0.013). Conclusions: Increased alterations in oncogenic drivers and estimates of CD8+ T cells and B cells were observed in the elderly population with HCC. The enhanced presence of co-inhibitory molecules suggests potential immune evasion. While we observed reduced TOT with sorafenib, additional studies are needed to elucidate the impact of molecular alterations on outcomes with sorafenib and newer therapies (i.e. immunotherapy) in older adults.

598 Background: PDAC of the H and B/T differ in embryonic origin, cell composition, blood supply, lymphatic and venous drainage, and innervation. H tumors tend cause symptoms earlier and to present at earlier stages compared to B/T cancers. The impact of PDAC tumor location on patient presentation and survival has been shown in large national data-based analyses, although with conflicting results. We aimed to compare the molecular and tumor immune microenvironment (TIME) profiles of PDAC of the H vs. B/T. Methods: A total of 3499 PDAC samples were analyzed via next-generation sequencing (NGS) of RNA (whole transcriptome, NovaSeq), DNA (NextSeq, 592 genes or NovaSeq, whole exome sequencing) and immunohistochemistry (IHC, Caris Life Sciences, Phoenix, AZ). RNA deconvolution was performed using QuantiSeq (Finotello 2019, Genome Medicine) to quantify the immune cell infiltration. Pathway gene enrichment analyses were done using Gene Set Enrichment Analysis (GSEA, Subramaniam 2015, PNAS). Significance was determined as p values adjusted for multiple correction (q) of < 0.05. Results: Anatomic subsites of PDAC tumors were grouped by primary tumor sites into H (N = 2058) or B/T (N = 1384). There were significantly more metastatic tumors profiled from H vs. B/T (57% vs. 44%, p < 0.001). KRAS mutations (93.8% vs. 90.2%), genomic loss of heterozygosity (12.7% vs. 9.1%), and several copy number alterations ( FGF3, FGF4, FGF19, CCND1, ZNF703, FLT4, MUTYH, TNFRS14) trended higher in B/T when compared to H (p < 0.05 but q > 0.05). GNAS mutations (2.2% vs. 0.7%) trended higher in H vs. B/T (p < 0.05). No significant difference in immuno-oncology (IO) markers (TMB, PD-L1, MSI-H) were observed, but expression analysis of IO-related genes showed significantly higher expression of CTLA4 and PDCD1 in H (q < 0.05, fold change 1.2 and 1.3) and IDO1 and PDCD1LG2 expression trended higher in B/T (p < 0.05, fold change 0.95). When comparing median cell abundance values as part of TIME analysis, H had increased immune infiltration of B cells (0.045 vs. 0.043), M2 macrophages (0.035 vs. 0.032), neutrophils (0.056 vs. 0.052), NK cells (0.027 vs. 0.026), CD8+ T cells (% > 0: 48.2% vs. 43.2%), while B/T had increased infiltration of M1 macrophages (0.035 vs. 0.032) (all q < 0.05). GSEA showed enrichment of CTLA4 (normalized enrichment score (NES) 1.6, false discovery rate (FDR) 0.19) and primary immunodeficiency pathway enrichment (NES 1.7, FDR 0.11) in H. Conclusions: To our knowledge, this is one of the largest cohort of PDAC tumors subjected to broad molecular profiling. Differences in IO-related gene expression and TIME cell distribution suggest that response to IO therapies may differ in PDAC arising from H vs B/T. Subtle differences in the genomic profliles of H vs. B/T tumors were also observed.

Use of Unmanned Aerial Systems (UASs) has grown steadily since the 1990s. They are becoming a standard tool for the military and are growing more commonplace in civilian applications. UASs of a vast array of sizes, capabilities and functions, are increasingly present in the National Airspace (NAS) and pressure to normalize their operations is taking on increasing urgency. The human interface for such systems will remain a critical component—and may increase in importance as human interactions are increasingly distanced, in time and/or space, from the locus of execution. Of particular complexity and concern are the use of larger UASs in dense traffic areas (such as cities) under high autonomy and with multiple vehicles per human operator. This paper summarizes a recently-completed review of multiple UAS operational concepts and evaluation of a suite of user interaction approaches for them. We have paid special attention to enabling ongoing situation awareness and prompt context switching. We have extensively leveraged our prior work on "Playbooks" for human tasking and delegation to automation. We conclude by presenting three novel human-machine interaction tools or approaches for such multi-UAS, high complexity, "routine" operations in the NAS: (1) a "strip-chart" view for combining vehicle-centric and task- or function-centric information in a timeline view, (2) an Activity Interchange display for facilitating task and vehicle handoffs within a defined task or play grammar, and (3) an approach applying plan recognition to verbal interactions to enable more intelligent and context-aware speech interactions for UAS control to speed and reduce workload in operator, Air Traffic Control and UAS interactions.

Recent advances in technology are delivering robots of reduced size and cost. A natural outgrowth of these advances are systems comprised of large numbers of robots that collaborate autonomously in diverse applications. Research on effective autonomous control of such systems, commonly called swarms, has increased dramatically in recent years and received attention from many domains, such as bioinspired robotics and control theory. These kinds of distributed systems present novel challenges for the effective integration of human supervisors, operators, and teammates that are only beginning to be addressed. This paper is the first survey of human–swarm interaction (HSI) and identifies the core concepts needed to design a human–swarm system. We first present the basics of swarm robotics. Then, we introduce HSI from the perspective of a human operator by discussing the cognitive complexity of solving tasks with swarm systems. Next, we introduce the interface between swarm and operator and identify challenges and solutions relating to human–swarm communication, state estimation and visualization, and human control of swarms. For the latter, we develop a taxonomy of control methods that enable operators to control swarms effectively. Finally, we synthesize the results to highlight remaining challenges, unanswered questions, and open problems for HSI, as well as how to address them in future works.

As robots become increasingly common in a wide variety of domains—from military and scientific applications to entertainment and home use—there is an increasing need to define and assess the trust humans have when interacting with robots. In human interaction with robots and automation, previous work has discovered that humans often have a tendency to either overuse automation, especially in cases of high workload, or underuse automation, both of which can make negative outcomes more likely. Frthermore, this is not limited to naive users, but experienced ones as well. Robotics brings a new dimension to previous work in trust in automation, as they are envisioned by many to work as teammates with their operators in increasingly complex tasks. In this chapter, our goal is to highlight previous work in trust in automation and human-robot interaction and draw conclusions and recommendations based on the existing literature. We believe that, while significant progress has been made in recent years, especially in quantifying and modeling trust, there are still several places where more investigation is needed.

In this paper, we study the model of human trust where an operator controls a robotic swarm remotely for a search mission. Existing trust models in human-in-the-loop systems are based on task performance of robots. However, we find that humans tend to make their decisions based on physical characteristics of the swarm rather than its performance since task performance of swarms is not clearly perceivable by humans. We formulate trust as a Markov decision process whose state space includes physical parameters of the swarm. We employ an inverse reinforcement learning algorithm to learn behaviors of the operator from a single demonstration. The learned behaviors are used to predict the trust level of the operator based on the features of the swarm.