Search

Significance: Eliminating "persisters" before relapse is crucial for achieving durable treatment efficacy. This study provides a rationale for developing PRMT1-selective inhibitors to target cancer persisters and achieve more durable outcomes in oncogene-targeting therapies., (©2024 The Authors; Published by the American Association for Cancer Research.)

Amyotrophic lateral sclerosis (ALS) is a rapidly progressing, highly heterogeneous neurodegenerative disease, underscoring the importance of obtaining information to personalize clinical decisions quickly after diagnosis. Here, we investigated whether ALS-relevant signatures can be detected directly from biopsied patient fibroblasts. We profiled familial ALS (fALS) fibroblasts, representing a range of mutations in the fused in sarcoma (FUS) gene and ages of onset. To differentiate FUS fALS and healthy control fibroblasts, machine-learning classifiers were trained separately on high-content imaging and transcriptional profiles. "Molecular ALS phenotype" scores, derived from these classifiers, captured a spectrum from disease to health. Interestingly, these scores negatively correlated with age of onset, identified several pre-symptomatic individuals and sporadic ALS (sALS) patients with FUS-like fibroblasts, and quantified "movement" of FUS fALS and "FUS-like" sALS toward health upon FUS ASO treatment. Taken together, these findings provide evidence that non-neuronal patient fibroblasts can be used for rapid, personalized assessment in ALS., Competing Interests: Declaration of interests L.F.W., M.P.J., and S.J.A. are founders and scientific advisory board members of Nine Square Therapeutics. L.F.W. and S.J.A. are advisory members of Developmental Cell’s advisory board., (Copyright © 2024 The Authors. Published by Elsevier Inc. All rights reserved.)

Importance: Addressing poor uptake of low-dose computed tomography lung cancer screening (LCS) is critical, especially for those having the most to gain-high-benefit persons with high lung cancer risk and life expectancy more than 10 years., Objective: To assess the association between LCS uptake and implementing a prediction-augmented shared decision-making (SDM) tool, which enables clinicians to identify persons predicted to be at high benefit and encourage LCS more strongly for these persons., Design, Setting, and Participants: Quality improvement interrupted time series study at 6 Veterans Affairs sites that used a standard set of clinical reminders to prompt primary care clinicians and screening coordinators to engage in SDM for LCS-eligible persons. Participants were persons without a history of LCS who met LCS eligibility criteria at the time (aged 55-80 years, smoked ≥30 pack-years, and current smoking or quit <15 years ago) and were not documented to be an inappropriate candidate for LCS by a clinician during October 2017 through September 2019. Data were analyzed from September to November 2023., Exposure: Decision support tool augmented by a prediction model that helps clinicians personalize SDM for LCS, tailoring the strength of screening encouragement according to predicted benefit., Main Outcome and Measure: LCS uptake., Results: In a cohort of 9904 individuals, the median (IQR) age was 64 (57-69) years; 9277 (94%) were male, 1537 (16%) were Black, 8159 (82%) were White, 5153 (52%) were predicted to be at intermediate (preference-sensitive) benefit and 4751 (48%) at high benefit, and 1084 (11%) received screening during the study period. Following implementation of the tool, higher rates of LCS uptake were observed overall along with an increase in benefit-based LCS uptake (higher screening uptake among persons anticipated to be at high benefit compared with those at intermediate benefit; primary analysis). Mean (SD) predicted probability of getting screened for a high-benefit person was 24.8% (15.5%) vs 15.8% (11.8%) for a person at intermediate benefit (mean absolute difference 9.0 percentage points; 95% CI, 1.6%-16.5%)., Conclusions and Relevance: Implementing a robust approach to personalized LCS, which integrates SDM, and a decision support tool augmented by a prediction model, are associated with improved uptake of LCS and may be particularly important for those most likely to benefit. These findings are timely given the ongoing poor rates of LCS uptake.

Introduction: Metastatic castration-resistant prostate cancer (mCRPC) is a heterogeneous disease with prognoses varying from months to years at time of castration-resistant diagnosis. Optimal first-line therapy for those with different prognoses is unknown., Methods: We conducted a retrospective cohort study of men in a national healthcare delivery system receiving first-line therapy for mCRPC (abiraterone, enzalutamide, docetaxel, or ketoconazole) from 2010 to 2017, with follow-up through 2019. Using commonly drawn prognostic labs at start of mCRPC therapy (hemoglobin, albumin, and alkaline phosphatase), we categorized men into favorable, intermediate, or poor prognostic groups depending on whether they had none, one to two, or all three laboratory values worse than designated laboratory cutoffs. We used Kaplan-Meier methods to examine prostate specific antigen (PSA) progression-free and overall survival (OS) according to prognostic group and first-line therapy, and multivariable cox regression to determine variables associated with survival outcomes., Results: Among 4135 patients, median PSA progression-free survival (PFS) was 6.9 months (95% confidence interval [CI] 6.6-7.3), and median OS 18.8 months (95% CI 18.0-19.6), ranging from 5.7 months (95% CI 4.8-7.0) in the poor prognosis group to 31.3 months (95% CI 29.7-32.9) in the favorable group. OS was similar regardless of initial treatment received for favorable and intermediate groups, but worse for those in the poor prognostic group who received ketoconazole (adjusted hazard ratio 2.07, 95% CI 1.2-3.6). PSA PFS was worse for those who received ketoconazole compared to abiraterone across all prognostic groups (favorable HR 1.76, 95% CI 1.34-2.31; intermediate HR 1.78, 95% CI 1.41-2.25; poor HR 8.01, 95% CI 2.93-21.9)., Conclusion: Commonly drawn labs at mCRPC treatment start may aid in predicting survival and response to therapies, potentially informing discussions with care teams. First-line treatment selection impacts disease progression for all men with mCRPC regardless of prognostic group, but impacted OS only for men with poor prognosis at treatment start., (© 2024 The Author(s). Cancer Medicine published by John Wiley & Sons Ltd.)

The combinatorial effect of genetic variants is often assumed to be additive. Although genetic variation can clearly interact non-additively, methods to uncover epistatic relationships remain in their infancy. We develop low-signal signed iterative random forests to elucidate the complex genetic architecture of cardiac hypertrophy. We derive deep learning-based estimates of left ventricular mass from the cardiac MRI scans of 29,661 individuals enrolled in the UK Biobank. We report epistatic genetic variation including variants close to CCDC141 , IGF1R , TTN , and TNKS. Several loci where variants were deemed insignificant in univariate genome-wide association analyses are identified. Functional genomic and integrative enrichment analyses reveal a complex gene regulatory network in which genes mapped from these loci share biological processes and myogenic regulatory factors. Through a network analysis of transcriptomic data from 313 explanted human hearts, we found strong gene co-expression correlations between these statistical epistasis contributors in healthy hearts and a significant connectivity decrease in failing hearts. We assess causality of epistatic effects via RNA silencing of gene-gene interactions in human induced pluripotent stem cell-derived cardiomyocytes. Finally, single-cell morphology analysis using a novel high-throughput microfluidic system shows that cardiomyocyte hypertrophy is non-additively modifiable by specific pairwise interactions between CCDC141 and both TTN and IGF1R . Our results expand the scope of genetic regulation of cardiac structure to epistasis.

Detecting epistatic drivers of human phenotypes is a considerable challenge. Traditional approaches use regression to sequentially test multiplicative interaction terms involving pairs of genetic variants. For higher-order interactions and genome-wide large-scale data, this strategy is computationally intractable. Moreover, multiplicative terms used in regression modeling may not capture the form of biological interactions. Building on the Predictability, Computability, Stability (PCS) framework, we introduce the epiTree pipeline to extract higher-order interactions from genomic data using tree-based models. The epiTree pipeline first selects a set of variants derived from tissue-specific estimates of gene expression. Next, it uses iterative random forests (iRF) to search training data for candidate Boolean interactions (pairwise and higher-order). We derive significance tests for interactions, based on a stabilized likelihood ratio test, by simulating Boolean tree-structured null (no epistasis) and alternative (epistasis) distributions on hold-out test data. Finally, our pipeline computes PCS epistasis p-values that probabilisticly quantify improvement in prediction accuracy via bootstrap sampling on the test set. We validate the epiTree pipeline in two case studies using data from the UK Biobank: predicting red hair and multiple sclerosis (MS). In the case of predicting red hair, epiTree recovers known epistatic interactions surrounding MC1R and novel interactions, representing non-linearities not captured by logistic regression models. In the case of predicting MS, a more complex phenotype than red hair, epiTree rankings prioritize novel interactions surrounding HLA-DRB1, a variant previously associated with MS in several populations. Taken together, these results highlight the potential for epiTree rankings to help reduce the design space for follow up experiments., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2024 Behr et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

The combinatorial effect of genetic variants is often assumed to be additive. Although genetic variation can clearly interact non-additively, methods to uncover epistatic relationships remain in their infancy. We develop low-signal signed iterative random forests to elucidate the complex genetic architecture of cardiac hypertrophy. We derive deep learning-based estimates of left ventricular mass from the cardiac MRI scans of 29,661 individuals enrolled in the UK Biobank. We report epistatic genetic variation including variants close to CCDC141 , IGF1R , TTN , and TNKS. Several loci not prioritized by univariate genome-wide association analysis are identified. Functional genomic and integrative enrichment analyses reveal a complex gene regulatory network in which genes mapped from these loci share biological processes and myogenic regulatory factors. Through a network analysis of transcriptomic data from 313 explanted human hearts, we show that these interactions are preserved at the level of the cardiac transcriptome. We assess causality of epistatic effects via RNA silencing of gene-gene interactions in human induced pluripotent stem cell-derived cardiomyocytes. Finally, single-cell morphology analysis using a novel high-throughput microfluidic system shows that cardiomyocyte hypertrophy is non-additively modifiable by specific pairwise interactions between CCDC141 and both TTN and IGF1R . Our results expand the scope of genetic regulation of cardiac structure to epistasis., Competing Interests: Competing interests E.A.A. is a Founder of Personalis, Deepcell, Svexa, RCD Co, and Parameter Health; Advisor to Oxford Nanopore, SequenceBio, and Pacific Biosciences; and a non-executive director for AstraZeneca. C.S.W. is a consultant for Tensixteen Bio and Renovacor. V.N.P. is an SAB member for and receives research support from BioMarin, Inc, and is a consultant for Constantiam, Inc. and viz.ai. The remaining authors declare no competing interests.

Background: Abiraterone and enzalutamide are castration-resistant prostate cancer (CRPC) therapies with potentially distinct associations with mental health symptoms given their differing antiandrogen targets., Methods: We used national Veterans Health Administration data to identify patients with CRPC who received first-line abiraterone or enzalutamide from 2010 to 2017. Using Poisson regression, we compared outpatient mental health encounters per 100 patient-months on drug between the abiraterone and enzalutamide cohorts adjusting for patient factors (e.g., age). We compared mental health encounters in the year before versus after starting therapy using the McNemar test., Results: We identified 2902 CRPC patients who received abiraterone (n = 1992) or enzalutamide (n = 910). We found no difference in outpatient mental health encounters between the two groups (adjusted incident rate ratio [aIRR] 1.04, 95% confidence interval [CI] 0.95-1.15). However, men with preexisting mental health diagnoses received 81.3% of the outpatient mental health encounters and had higher rates of these encounters with enzalutamide (aIRR 1.21, 95% CI 1.09-1.34). Among patients with ≥1 year of enrollment before and after starting abiraterone (n = 1139) or enzalutamide (n = 446), there was no difference in mental health care utilization before versus after starting treatment (17.0% of patients vs. 17.6%, p = 0.60, abiraterone; 16.4% vs. 18.4%, p = 0.26, enzalutamide)., Conclusion: We found no overall differences in mental health care utilization between CRPC patients who received first-line abiraterone versus enzalutamide. However, men with preexisting mental health diagnoses received the majority of mental health care and had more mental health visits with enzalutamide., (© 2023 The Authors. Cancer Medicine published by John Wiley & Sons Ltd.)

Parkinson's disease-causing leucine-rich repeat kinase 2 (LRRK2) mutations lead to varying degrees of Rab GTPase hyperphosphorylation. Puzzlingly, LRRK2 GTPase-inactivating mutations-which do not affect intrinsic kinase activity-lead to higher levels of cellular Rab phosphorylation than kinase-activating mutations. Here, we investigate whether mutation-dependent differences in LRRK2 cellular localization could explain this discrepancy. We discover that blocking endosomal maturation leads to the rapid formation of mutant LRRK2 + endosomes on which LRRK2 phosphorylates substrate Rabs. LRRK2 + endosomes are maintained through positive feedback, which mutually reinforces membrane localization of LRRK2 and phosphorylated Rab substrates. Furthermore, across a panel of mutants, cells expressing GTPase-inactivating mutants form strikingly more LRRK2 + endosomes than cells expressing kinase-activating mutants, resulting in higher total cellular levels of phosphorylated Rabs. Our study suggests that the increased probability that LRRK2 GTPase-inactivating mutants are retained on intracellular membranes compared to kinase-activating mutants leads to higher substrate phosphorylation., Competing Interests: Declaration of interests L.F.W., M.P.J., and S.J.A. are founders, SAB members, and paid consultants of Nine Square Therapeutics., (Copyright © 2023 The Author(s). Published by Elsevier Inc. All rights reserved.)

Objectives: To determine the association between covid-19 vaccination types and doses with adverse outcomes of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection during the periods of delta (B.1.617.2) and omicron (B.1.1.529) variant predominance., Design: Retrospective cohort., Setting: US Veterans Affairs healthcare system., Participants: Adults (≥18 years) who are affiliated to Veterans Affairs with a first documented SARS-CoV-2 infection during the periods of delta (1 July-30 November 2021) or omicron (1 January-30 June 2022) variant predominance. The combined cohorts had a mean age of 59.4 (standard deviation 16.3) and 87% were male., Interventions: Covid-19 vaccination with mRNA vaccines (BNT162b2 (Pfizer-BioNTech) and mRNA-1273 (Moderna)) and adenovirus vector vaccine (Ad26.COV2.S (Janssen/Johnson & Johnson))., Main Outcome Measures: Stay in hospital, intensive care unit admission, use of ventilation, and mortality measured 30 days after a positive test result for SARS-CoV-2., Results: In the delta period, 95 336 patients had infections with 47.6% having at least one vaccine dose, compared with 184 653 patients in the omicron period, with 72.6% vaccinated. After adjustment for patient demographic and clinical characteristics, in the delta period, two doses of the mRNA vaccines were associated with lower odds of hospital admission (adjusted odds ratio 0.41 (95% confidence interval 0.39 to 0.43)), intensive care unit admission (0.33 (0.31 to 0.36)), ventilation (0.27 (0.24 to 0.30)), and death (0.21 (0.19 to 0.23)), compared with no vaccination. In the omicron period, receipt of two mRNA doses were associated with lower odds of hospital admission (0.60 (0.57 to 0.63)), intensive care unit admission (0.57 (0.53 to 0.62)), ventilation (0.59 (0.51 to 0.67)), and death (0.43 (0.39 to 0.48)). Additionally, a third mRNA dose was associated with lower odds of all outcomes compared with two doses: hospital admission (0.65 (0.63 to 0.69)), intensive care unit admission (0.65 (0.59 to 0.70)), ventilation (0.70 (0.61 to 0.80)), and death (0.51 (0.46 to 0.57)). The Ad26.COV2.S vaccination was associated with better outcomes relative to no vaccination, but higher odds of hospital stay and intensive care unit admission than with two mRNA doses. BNT162b2 was generally associated with worse outcomes than mRNA-1273 (adjusted odds ratios between 0.97 and 1.42)., Conclusions: In veterans with recent healthcare use and high occurrence of multimorbidity, vaccination was robustly associated with lower odds of 30 day morbidity and mortality compared with no vaccination among patients infected with covid-19. The vaccination type and number of doses had a significant association with outcomes., Competing Interests: Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/disclosure-of-interest/ and declare support from the US Department of Veterans Affairs for the submitted work but no financial relationship with any organisation that might have an interest in the submitted work in the previous three years and no other relationships or activities that could appear to have influenced the submitted work., (© Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.)

High-content microscopy offers a scalable approach to screen against multiple targets in a single pass. Prior work has focused on methods to select "optimal" cellular readouts in microscopy screens. However, methods to select optimal cell line models have garnered much less attention. Here, we provide a roadmap for how to select the cell line or lines that are best suited to identify bioactive compounds and their mechanism of action (MOA). We test our approach on compounds targeting cancer-relevant pathways, ranking cell lines in two tasks: detecting compound activity ("phenoactivity") and grouping compounds with similar MOA by similar phenotype ("phenosimilarity"). Evaluating six cell lines across 3214 well-annotated compounds, we show that optimal cell line selection depends on both the task of interest (e.g., detecting phenoactivity vs inferring phenosimilarity) and distribution of MOAs within the compound library. Given a task of interest and a set of compounds, we provide a systematic framework for choosing optimal cell line(s). Our framework can be used to reduce the number of cell lines required to identify hits within a compound library and help accelerate the pace of early drug discovery.

Background: Over the past decade, abiraterone and enzalutamide have largely replaced ketoconazole as oral treatments for castration-resistant prostate cancer (CRPC). We investigated the differential adoption of abiraterone and enzalutamide across facilities in a national healthcare system to understand the impact a facility has on the receipt of these novel therapies., Methods: Using data from the VA Corporate Data Warehouse, we identified a cohort of men with CRPC who received the most common first-line therapies: abiraterone, enzalutamide, docetaxel, or ketoconazole between 2010 and 2017. We described variability in the adoption of abiraterone and enzalutamide across facilities by time period (2010-2013 or 2014-2017). We categorized facilities depending on the timing of adoption of abiraterone and enzalutamide relative to other facilities and described facility characteristics associated with early and late adoption., Results: We identified 4998 men treated with ketoconazole, docetaxel, abiraterone, or enzalutamide as first-line CRPC therapy between 2010 and 2017 at 125 national facilities. When limiting the cohort to oral therapies, most patients treated earlier in the study period (2010-2013) received ketoconazole. A dramatic shift was seen by the second half of the study period (2014-2017) with most men treated with first-line abiraterone (61%). Despite this shift and a new standard of care, some facilities persisted in the widespread use of ketoconazole in the later period, so-called late adopting facilities. After multivariable adjustment, patients who received treatment at a late adopting facility were more likely receiving care at a lower complexity, rural facility, with less urology and hematology/oncology workforce (all p < 0.01)., Conclusion: Many facilities persisted in their use of ketoconazole as first-line CRPC therapy, even when other facilities had adopted the new standard of care abiraterone and enzalutamide. Further work is needed to identify the effect of this late adoption on outcomes important to patients., (© 2022 The Authors. Cancer Medicine published by John Wiley & Sons Ltd.)

Increasing occurrence of harmful algal blooms across the land-water interface poses significant risks to coastal ecosystem structure and human health. Defining significant drivers and their interactive impacts on blooms allows for more effective analysis and identification of specific conditions supporting phytoplankton growth. A novel iterative Random Forests (iRF) machine-learning model was developed and applied to two example cases along the California coast to identify key stable interactions: (1) phytoplankton abundance in response to various drivers due to coastal conditions and land-sea nutrient fluxes, (2) microbial community structure during algal blooms. In Example 1, watershed derived nutrients were identified as the least significant interacting variable associated with Monterey Bay phytoplankton abundance. In Example 2, through iRF analysis of field-based 16S OTU bacterial community and algae datasets, we independently found stable interactions of prokaryote abundance patterns associated with phytoplankton abundance that have been previously identified in laboratory-based studies. Our study represents the first iRF application to marine algal blooms that helps to identify ocean, microbial, and terrestrial conditions that are considered dominant causal factors on bloom dynamics., (© 2021. The Author(s).)

Background: Treatments for metastatic castration-resistant prostate cancer (CRPC) differ in toxicity, administration, and evidence. In this study, clinical and nonclinical factors associated with the first-line treatment for CRPC in a national delivery system were evaluated., Methods: National electronic laboratory and clinical data from the Veterans Health Administration were used to identify patients with CRPC (ie, rising prostate-specific antigen [PSA] on androgen deprivation) who received abiraterone, enzalutamide, docetaxel, or ketoconazole from 2010 through 2017. It was determined whether clinical (eg, PSA) and nonclinical factors (eg, race, facility) were associated with the first-line treatment selection using multilevel, multinomial logistic regression. The average marginal effects (AMEs) were calculated of patient, disease, and facility characteristics on ketoconazole versus more appropriate CRPC therapy., Results: There were 4998 patients identified with CRPC who received first-line ketoconazole, docetaxel, abiraterone, or enzalutamide. After adjustment, increasing age was associated with receipt of abiraterone (adjusted odds ratio [aOR], 1.07; 95% credible interval [CrI], 1.06-1.09) or enzalutamide (aOR, 1.10; 95% CrI, 1.08-1.11) versus docetaxel. Greater preexisting comorbidity was associated with enzalutamide versus abiraterone (aOR, 1.53; 95% CrI, 1.23-1.91). Patients with higher PSA values at the start of treatment were more likely to receive docetaxel than oral agents and less likely to receive ketoconazole than other oral agents. African American men were more likely to receive ketoconazole than abiraterone, enzalutamide, or docetaxel (AME, 2.8%; 95% CI, 0.7%-4.9%). This effect was attenuated when adjusting for facility characteristics (AME, 1.9%; 95% CI, -0.4% to 4.1%)., Conclusions: Clinical factors had an expected effect on the first-line treatment selection. Race may be associated with the receipt of a guideline-discordant first-line treatment., (© 2021 American Cancer Society.)

Building and expanding on principles of statistics, machine learning, and scientific inquiry, we propose the predictability, computability, and stability (PCS) framework for veridical data science. Our framework, composed of both a workflow and documentation, aims to provide responsible, reliable, reproducible, and transparent results across the data science life cycle. The PCS workflow uses predictability as a reality check and considers the importance of computation in data collection/storage and algorithm design. It augments predictability and computability with an overarching stability principle. Stability expands on statistical uncertainty considerations to assess how human judgment calls impact data results through data and model/algorithm perturbations. As part of the PCS workflow, we develop PCS inference procedures, namely PCS perturbation intervals and PCS hypothesis testing, to investigate the stability of data results relative to problem formulation, data cleaning, modeling decisions, and interpretations. We illustrate PCS inference through neuroscience and genomics projects of our own and others. Moreover, we demonstrate its favorable performance over existing methods in terms of receiver operating characteristic (ROC) curves in high-dimensional, sparse linear model simulations, including a wide range of misspecified models. Finally, we propose PCS documentation based on R Markdown or Jupyter Notebook, with publicly available, reproducible codes and narratives to back up human choices made throughout an analysis. The PCS workflow and documentation are demonstrated in a genomics case study available on Zenodo., Competing Interests: The authors declare no competing interest., (Copyright © 2020 the Author(s). Published by PNAS.)

Machine-learning models have demonstrated great success in learning complex patterns that enable them to make predictions about unobserved data. In addition to using models for prediction, the ability to interpret what a model has learned is receiving an increasing amount of attention. However, this increased focus has led to considerable confusion about the notion of interpretability. In particular, it is unclear how the wide array of proposed interpretation methods are related and what common concepts can be used to evaluate them. We aim to address these concerns by defining interpretability in the context of machine learning and introducing the predictive, descriptive, relevant (PDR) framework for discussing interpretations. The PDR framework provides 3 overarching desiderata for evaluation: predictive accuracy, descriptive accuracy, and relevancy, with relevancy judged relative to a human audience. Moreover, to help manage the deluge of interpretation methods, we introduce a categorization of existing techniques into model-based and post hoc categories, with subgroups including sparsity, modularity, and simulatability. To demonstrate how practitioners can use the PDR framework to evaluate and understand interpretations, we provide numerous real-world examples. These examples highlight the often underappreciated role played by human audiences in discussions of interpretability. Finally, based on our framework, we discuss limitations of existing methods and directions for future work. We hope that this work will provide a common vocabulary that will make it easier for both practitioners and researchers to discuss and choose from the full range of interpretation methods., Competing Interests: The authors declare no competing interest.

Genomics has revolutionized biology, enabling the interrogation of whole transcriptomes, genome-wide binding sites for proteins, and many other molecular processes. However, individual genomic assays measure elements that interact in vivo as components of larger molecular machines. Understanding how these high-order interactions drive gene expression presents a substantial statistical challenge. Building on random forests (RFs) and random intersection trees (RITs) and through extensive, biologically inspired simulations, we developed the iterative random forest algorithm (iRF). iRF trains a feature-weighted ensemble of decision trees to detect stable, high-order interactions with the same order of computational cost as the RF. We demonstrate the utility of iRF for high-order interaction discovery in two prediction problems: enhancer activity in the early Drosophila embryo and alternative splicing of primary transcripts in human-derived cell lines. In Drosophila , among the 20 pairwise transcription factor interactions iRF identifies as stable (returned in more than half of bootstrap replicates), 80% have been previously reported as physical interactions. Moreover, third-order interactions, e.g., between Zelda ( Zld ), Giant ( Gt ), and Twist ( Twi ), suggest high-order relationships that are candidates for follow-up experiments. In human-derived cells, iRF rediscovered a central role of H3K36me3 in chromatin-mediated splicing regulation and identified interesting fifth- and sixth-order interactions, indicative of multivalent nucleosomes with specific roles in splicing regulation. By decoupling the order of interactions from the computational cost of identification, iRF opens additional avenues of inquiry into the molecular mechanisms underlying genome biology., Competing Interests: The authors declare no conflict of interest., (Copyright © 2018 the Author(s). Published by PNAS.)