Search

43 Background: Blood-based screening tests for colorectal cancer (CRC) with high sensitivity and specificity are needed to improve adherence, facilitate early detection, and ultimately reduce mortality from CRC. Current stool-based tests have a sensitivity of 24-42% for colorectal advanced adenomas (AAs), while blood tests that rely on tumor-derived cell-free DNA (cfDNA) methylation signatures have shown limited sensitivity for AAs. Here we demonstrate the ability to detect AAs from blood using a multiomics test that incorporates both tumor- and immune-derived signatures, and compare it to the performance of a cfDNA methylation-only test. Methods: Participants enrolled in a prospective study (NCT03688906) were included in this analysis. The multiomics test includes signatures for cell-free nucleic acids based on next-generation sequencing, and for plasma proteins based on high-throughput multiplexed assays. Signatures are integrated computationally with a combination of convolutional neural networks and regularized logistic regression. We compared the multiomics test with one based on cfDNA methylation only. Results: This sub-study included 542 participants (AA: n = 122; colonoscopy-confirmed negative controls: n = 420). Participants with AA were 56% male with a mean age of 63 years, and colonoscopy-confirmed negative controls were 54% male with a mean age of 61 years. The multiomics test achieved a sensitivity of 41% (n = 50/122, 95% CI 34-48%) at 90% specificity (377/420). By contrast, the cfDNA methylation-only test achieved a sensitivity of 20% (24/122, 95% CI 15-25%) at 91% specificity (383/420). Performance was also analyzed by histological subtype and location, and superiority of the multiomics test to the cfDNA-methylation-only test was consistently observed. Conclusions: A novel multiomics blood test can detect colorectal AAs at a sensitivity and specificity comparable to existing stool-based tests. Combining signatures from both tumor- and immune-derived sources resulted in AA sensitivity greater than that of cfDNA-methylation alone.

BackgroundBlood-based methods using cell-free DNA (cfDNA) are under development as an alternative to existing screening tests. However, early-stage detection of cancer using tumor-derived cfDNA has proven challenging because of the small proportion of cfDNA derived from tumor tissue in early-stage disease. A machine learning approach to discover signatures in cfDNA, potentially reflective of both tumor and non-tumor contributions, may represent a promising direction for the early detection of cancer.MethodsWhole-genome sequencing was performed on cfDNA extracted from plasma samples (N=546 colorectal cancer and 271 non-cancer controls). Reads aligning to protein-coding gene bodies were extracted, and read counts were normalized. cfDNA tumor fraction was estimated using IchorCNA. Machine learning models were trained using k-fold cross-validation and confounder-based cross-validation to assess generalization performance.ResultsIn a colorectal cancer cohort heavily weighted towards early-stage cancer (80% stage I/II), we achieved a mean AUC of 0.92 (95% CI 0.91-0.93) with a mean sensitivity of 85% (95% CI 83-86%) at 85% specificity. Sensitivity generally increased with tumor stage and increasing tumor fraction. Stratification by age, sequencing batch, and institution demonstrated the impact of these confounders and provided a more accurate assessment of generalization performance.ConclusionsA machine learning approach using cfDNA achieved high sensitivity and specificity in a large, predominantly early-stage, colorectal cancer cohort. The possibility of systematic technical and institution-specific biases warrants similar confounder analyses in other studies. Prospective validation of this machine learning method and evaluation of a multi-analyte approach are underway.

9588 Background: Immune checkpoint inhibitors have shown promising results in many advanced cancers, but the response rate remains low. Various molecular and cellular biomarkers, such as elevated tumor-infiltrating cytotoxic T cells and Natural Killer (NK) cells at baseline, are associated with response. Blood-based biomarkers to predict or monitor response remain challenging to develop. Here we investigate the potential of cell-free DNA (cfDNA) biomarkers to predict response to the PD-1 immune checkpoint inhibitor nivolumab in patients with refractory metastatic non-small cell lung cancer (NSCLC). Methods: Plasma from stage IV NSCLC patients enrolled in ALCINA (NCT02866149) was collected before (baseline, BL, n = 30) and at week 8 (W8, n = 17) of nivolumab therapy. Response was determined using RECIST 1.1 (responders n = 5; non-responders n = 25). Whole-genome sequencing was performed to characterize cfDNA fragments. Tumor fraction (TF) was assessed using ichorCNA. Cellular composition was estimated by deconvolution of cfDNA co-fragmentation patterns, and transcription factor activity was estimated by measuring binding site accessibility across the genome. Results: Although estimated TF at baseline did not predict response to nivolumab, NK cell levels estimated by cell-mixture deconvolution were significantly higher in responders at BL (p < 0.05). Furthermore, estimated monocyte levels at W8 strongly correlated with overall survival (r = 0.75, p < 0.0005, HR = 15.02) and were significantly higher in responders (p < 0.05). By evaluating changes in transcription factor binding activity, we identified factors with greater accessibility in non-responders at baseline (DEAF1, THAP11) and W8 (DUX4, PDX-1). Conclusions: Plasma cfDNA signatures may be useful for response prediction and monitoring in NSCLC patients on immunotherapy. Our results suggest that changes in the immune system, as reflected by cellular composition and transcriptional activity inferred from cfDNA, may provide biological insights beyond TF alone that may benefit biomarker discovery and drug target identification.

66 Background: Despite population screening efforts, screening rates for colorectal cancer (CRC) remain suboptimal. A non-invasive, blood-based screening test with high sensitivity and specificity in early-stage disease should improve adherence and ultimately reduce mortality; however, tests based only on tumor-derived biomarkers have limited sensitivity. Here we used a multiomic, machine learning platform to discover, refine, and combine tumor- and immune-derived signals to develop a blood test for the detection of early-stage CRC. Methods: Samples from 591 participants enrolled in a prospective study including average-risk screening and case-control cohorts (NCT03688906) were included in this analysis (CRC: n = 43; colonoscopy-confirmed CRC-negative controls: n = 548). Participants with CRC were 60% male with a mean age of 63, and controls were 55% male with a mean age of 60. Stage distribution was 54% early (I/II) and 34% late (III/IV) with 11% unknown. Plasma was analyzed by whole-genome sequencing, bisulfite sequencing, and protein quantification methods. Computational methods were used to assess and infer the performance of individual and combined assays. Results: For colorectal adenocarcinoma, which represents ~95% of all CRCs, our multiomic test achieved a mean sensitivity of 92% in early stage (n = 17) and 84% in late stage (n = 11) at a specificity of 90%. Across all CRC pathological subtypes, our test achieved a mean sensitivity of 80% in early stage (n = 19) and 83% in late stage (n = 12) at a specificity of 90%; the test detected the single squamous cell carcinoma but missed both neuroendocrine tumors. Individual assays achieved a mean sensitivity of 50% in early stage and 66% in late stage at a specificity of 90%. Conclusions: In a prospective cohort, we demonstrated high sensitivity and specificity for early-stage adenocarcinoma by combining tumor- and immune-derived signals from cfDNA, epigenetic, and protein biomarkers. While most CRCs are adenocarcinomas, detection of all pathological subtypes is required to maximize sensitivity in a screening population. Further analysis of molecular and pathological subtypes, as well as the entire ~3000 patient cohort, is underway. Clinical trial information: NCT03688906.

Introduction: Blood-based tests hold great promise as cancer diagnostics but until now have largely been restricted to the analysis of a single class of molecules (eg, circulating tumor DNA, platelet mRNA, circulating proteins). The ability to analyze multiple analytes simultaneously from the same biological sample may increase the sensitivity and specificity of such tests by exploiting independent information between signals. Here, we describe an experimental and analytical system that we developed and implemented for the integrated analysis of multiple analytes from a single blood sample, which revealed examples of both correlations and orthogonality among individual analytes. Methods: De-identified blood samples were obtained from healthy individuals, as well as individuals with pre-malignant conditions and stage I-IV colorectal cancer (CRC). After plasma separation, multiple types of analytes were assayed: cell-free DNA (cfDNA) content was assessed by low-coverage whole-genome sequencing (lcWGS) and whole-genome bisulfite sequencing (WGBS), cell-free microRNA (cf-miRNA) was assessed by small-RNA sequencing, and levels of circulating proteins were measured by quantitative immunoassay. Results: lcWGS of plasma cfDNA was able to identify CRC samples with high tumor fraction (>20%) on the basis of copy number variation (CNV) across the genome. High tumor fractions, while more frequent in late-stage cancer samples, were observed in some stage I and II patients. Aberrant signals in each of the three other analytes—cf-miRNA profiles discordant with those in healthy controls, genome-wide hypomethylation at LINE1 (long interspersed nuclear element 1) CpG loci, and elevated levels of circulating carcinoembryonic antigen (CEA) and cytokeratin fragment 21-1 (CYFRA 21-1) proteins—were also observed in cancer patients. Strikingly, aberrant profiles across analytes were indicative of high tumor fraction (as estimated from cfDNA CNV), rather than cancer stage. Conclusion: Our data suggest that tumor fraction is correlated with cancer stage but has a large potential range, even in early stage samples. Previous literature on blood-based screens for detection of cancer has displayed discordance in the claimed ability of different single analytes to detect early stage cancer. Tumor fraction may be able to explain the historical disagreement, as we found that aberrant profiles among cf-miRNA, cfDNA methylation, and circulating protein levels were more strongly associated with high tumor fraction than with late stage. These findings suggest that some positive “early stage” detection results may in fact be “high tumor fraction” detection results. Our results further demonstrate that assaying multiple analytes from a single sample may enable the development of classifiers that are reliable at low tumor fraction and for detecting pre-malignant or early-stage disease. Citation Format: Daniel Delubac, Eric Ariazi, Jonathan Berliner, Adam Drake, John Dulin, Riley Ennis, Erik Gafni, Kate Niehaus, Gabriel Otte, Jennifer Pecson, Girish Putcha, Corey Schaninger, Aarushi Sharma, Mike Singer, Abraham Tzou, Jill Waters, David Weinberg, Brandon White, Imran S. Haque. Multi-analyte profiling reveals relationships among circulating biomarkers in colorectal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2227.