An Active Learning Framework Improves Tumor Variant Interpretation.
Ontology highlight
ABSTRACT: Significance
A novel machine learning approach predicts the impact of tumor mutations on cellular phenotypes, overcomes limited training data, minimizes costly functional validation, and advances efforts to implement cancer precision medicine.
SUBMITTER: Blee AM
PROVIDER: S-EPMC9357215 | biostudies-literature | 2022 Aug
REPOSITORIES: biostudies-literature
Publications
An Active Learning Framework Improves Tumor Variant Interpretation.
Blee Alexandra M AM Li Bian B Pecen Turner T Meiler Jens J Nagel Zachary D ZD Capra John A JA Chazin Walter J WJ
Cancer research 20220801 15
<h4>Significance</h4>A novel machine learning approach predicts the impact of tumor mutations on cellular phenotypes, overcomes limited training data, minimizes costly functional validation, and advances efforts to implement cancer precision medicine. ...[more]
Similar Datasets
Project description:The acceleration of DNA sequencing in samples from patients and population studies has resulted in extensive catalogues of human genetic variation, but the interpretation of rare genetic variants remains problematic. A notable example of this challenge is the existence of disruptive variants in dosage-sensitive disease genes, even in apparently healthy individuals. Here, by manual curation of putative loss-of-function (pLoF) variants in haploinsufficient disease genes in the Genome Aggregation Database (gnomAD)1, we show that one explanation for this paradox involves alternative splicing of mRNA, which allows exons of a gene to be expressed at varying levels across different cell types. Currently, no existing annotation tool systematically incorporates information about exon expression into the interpretation of variants. We develop a transcript-level annotation metric known as the 'proportion expressed across transcripts', which quantifies isoform expression for variants. We calculate this metric using 11,706 tissue samples from the Genotype Tissue Expression (GTEx) project2 and show that it can differentiate between weakly and highly evolutionarily conserved exons, a proxy for functional importance. We demonstrate that expression-based annotation selectively filters 22.8% of falsely annotated pLoF variants found in haploinsufficient disease genes in gnomAD, while removing less than 4% of high-confidence pathogenic variants in the same genes. Finally, we apply our expression filter to the analysis of de novo variants in patients with autism spectrum disorder and intellectual disability or developmental disorders to show that pLoF variants in weakly expressed regions have similar effect sizes to those of synonymous variants, whereas pLoF variants in highly expressed exons are most strongly enriched among cases. Our annotation is fast, flexible and generalizable, making it possible for any variant file to be annotated with any isoform expression dataset, and will be valuable for the genetic diagnosis of rare diseases, the analysis of rare variant burden in complex disorders, and the curation and prioritization of variants in recall-by-genotype studies.
| S-EPMC7334198 | biostudies-literature
Project description:PurposeAccurate discrimination of benign and pathogenic rare variation remains a priority for clinical genome interpretation. State-of-the-art machine learning variant prioritization tools are imprecise and ignore important parameters defining gene-disease relationships, e.g., distinct consequences of gain-of-function versus loss-of-function variants. We hypothesized that incorporating disease-specific information would improve tool performance.MethodsWe developed a disease-specific variant classifier, CardioBoost, that estimates the probability of pathogenicity for rare missense variants in inherited cardiomyopathies and arrhythmias. We assessed CardioBoost's ability to discriminate known pathogenic from benign variants, prioritize disease-associated variants, and stratify patient outcomes.ResultsCardioBoost has high global discrimination accuracy (precision recall area under the curve [AUC] 0.91 for cardiomyopathies; 0.96 for arrhythmias), outperforming existing tools (4-24% improvement). CardioBoost obtains excellent accuracy (cardiomyopathies 90.2%; arrhythmias 91.9%) for variants classified with >90% confidence, and increases the proportion of variants classified with high confidence more than twofold compared with existing tools. Variants classified as disease-causing are associated with both disease status and clinical severity, including a 21% increased risk (95% confidence interval [CI] 11-29%) of severe adverse outcomes by age 60 in patients with hypertrophic cardiomyopathy.ConclusionsA disease-specific variant classifier outperforms state-of-the-art genome-wide tools for rare missense variants in inherited cardiac conditions ( https://www.cardiodb.org/cardioboost/ ), highlighting broad opportunities for improved pathogenicity prediction through disease specificity.
| S-EPMC7790749 | biostudies-literature
Project description:As genomic sequencing expands, so does our knowledge of the link between genetic variation and disease. Deeper catalogs of variant frequencies improve identification of benign variants, while sequencing affected individuals reveals disease-associated variation. Accumulation of human genetic data thus makes reanalysis a means to maximize the benefits of clinical sequencing. We implemented pipelines to systematically reassess sequencing data from 494 individuals with developmental disability. Reanalysis yielded pathogenic or likely pathogenic (P/LP) variants that were not initially reported in 23 individuals, 6 described here, comprising a 16% increase in P/LP yield. We also downgraded 3 LP and 6 variants of uncertain significance (VUS) due to updated population frequency data. The likelihood of identifying a new P/LP variant increased over time, as ~22% of individuals who did not receive a P/LP variant at their original analysis subsequently did after 3 years. We show here that reanalysis and data sharing increase the diagnostic yield and accuracy of clinical sequencing.
| S-EPMC5995667 | biostudies-literature
Project description:PurposeAutosomal dominant polycystic kidney disease (ADPKD) is a common adult-onset monogenic disorder, with prevalence of 1/1000. Population databases including ExAC have improved pathogenic variant prioritization in many diseases. Due to pseudogene homology of PKD1, the predominant ADPKD disease gene, and the variable disease severity and age of onset, we aimed to investigate the utility of ExAC for variant assessment in ADPKD.MethodsWe assessed coverage and variant quality in the ExAC cohort and combined allele frequency and age data from the ExAC database (n = 60,706) with curated variants from 2000 ADPKD pedigrees (ADPKD Mutation Database).ResultsSeventy-six percent of PKD1 and PKD2 were sequenced adequately for variant discovery and variant quality was high in ExAC. In ExAC, we identified 25 truncating and 393 previously reported disease-causing variants in PKD1 and PKD2, 6.9-fold higher than expected. Fifty-four different variants, previously classified as disease-causing, were observed in ≥5 participants in ExAC.ConclusionOur study demonstrates that many previously implicated disease-causing variants are too common, challenging their pathogenicity, or penetrance. The presence of protein-truncating variants in older participants in ExAC demonstrates the complexity of variant classification and highlights need for further study of prevalence and penetrance of this common monogenic disease.
| S-EPMC6677146 | biostudies-literature
Project description:Microsatellite instability (MSI) is a critical phenotype of cancer genomes and an FDA-recognized biomarker that can guide treatment with immune checkpoint inhibitors. Previous work has demonstrated that next-generation sequencing data can be used to identify samples with MSI-high phenotype. However, low tumor purity, as frequently observed in routine clinical samples, poses a challenge to the sensitivity of existing algorithms. To overcome this critical issue, we developed MiMSI, an MSI classifier based on deep neural networks and trained using a dataset that included low tumor purity MSI cases in a multiple instance learning framework. On a challenging yet representative set of cases, MiMSI showed higher sensitivity (0.895) and auROC (0.971) than MSISensor (sensitivity: 0.67; auROC: 0.907), an open-source software previously validated for clinical use at our institution using MSK-IMPACT large panel targeted NGS data. In a separate, prospective cohort, MiMSI confirmed that it outperforms MSISensor in low purity cases (P = 8.244e-07).
| S-EPMC11696176 | biostudies-literature
Project description:IntroductionHead computed tomography (CT) interpretation is a vital skill for emergency physicians. Existing literature shows poor concordance between emergency physicians and radiologists in head CT interpretation. Prior studies have used passive learning methods to address this knowledge gap. We created an active learning curriculum for teaching head CT interpretation to emergency medicine (EM) residents and compared its effectiveness to a passive learning strategy.MethodsWe conducted a prospective, randomized controlled study of EM residents at a single institution. Three educational sessions were delivered over a three-month period via video conference. The active learning cohort (ALC) scrolled through head CT teaching cases we designed on Pascbin, a web-based radiology picture archiving and communication system. The passive learning cohort (PLC) watched instructional videos that scrolled through the same cases. Both cohorts were given equal time to review the cases and ask an instructor questions. Residents took pre-intervention and post-intervention tests on head CT interpretation. We analyzed scores using paired and unpaired t-tests.ResultsForty-two residents took the pre-intervention test. Mean pre- and post-test scores for the ALC were 43.8% and 59.0% (P <0.001), and for the PLC were 41.7% and 45.3% (P = 0.29). The difference in ALC and PLC post-test scores was statistically significant (P = 0.009) with a large effect size (Cohen's d = 1.34).ConclusionOur active learning head CT curriculum using Pacsbin showed superior learning outcomes when compared to a passive learning strategy and required no additional time or resources. This intervention offers a more effective and learner-centric method for implementing radiology curricula in EM residency programs.
| S-EPMC8782134 | biostudies-literature
Project description:Analysis of genomic DNA methylation by generating epigenetic signature profiles (episignatures) is increasingly being implemented in genetic diagnosis. Here we report our experience using episignature analysis to resolve both uncomplicated and complex cases of neurodevelopmental disorders (NDDs). We analyzed 97 NDDs divided into (1) a validation cohort of 59 patients with likely pathogenic/pathogenic variants characterized by a known episignature and (2) a test cohort of 38 patients harboring variants of unknown significance or unidentified variants. The expected episignature was obtained in most cases with likely pathogenic/pathogenic variants (53/59 [90%]), a revealing exception being the overlapping profile of two SMARCB1 pathogenic variants with ARID1A/B:c.6200, confirmed by the overlapping clinical features. In the test cohort, five cases showed the expected episignature, including (1) novel pathogenic variants in ARID1B and BRWD3; (2) a deletion in ATRX causing MRXFH1 X-linked mental retardation; and (3) confirmed the clinical diagnosis of Cornelia de Lange (CdL) syndrome in mutation-negative CdL patients. Episignatures analysis of the in BAF complex components revealed novel functional protein interactions and common episignatures affecting homologous residues in highly conserved paralogous proteins (SMARCA2 M856V and SMARCA4 M866V). Finally, we also found sex-dependent episignatures in X-linked disorders. Implementation of episignature profiling is still in its early days, but with increasing utilization comes increasing awareness of the capacity of this methodology to help resolve the complex challenges of genetic diagnoses.
| S-EPMC11216013 | biostudies-literature
Project description:Discovery of potential drugs requires rapid and precise identification of drug targets. Although traditional experimental methodologies can accurately identify drug targets, they are time-consuming and inappropriate for high-throughput screening. Computational approaches based on machine learning (ML) algorithms can expedite the prediction of druggable proteins; however, the performance of the existing computational methods remains unsatisfactory. This study proposes a computational tool, SPIDER, to enhance the accurate prediction of druggable proteins. SPIDER employs various feature descriptors pertaining to several aspects, including physicochemical properties, compositional information, and composition-transition-distribution information, coupled with well-known ML algorithms to facilitate the construction of the final meta-predictor. The experimental results showed that SPIDER enabled more precise and robust prediction of druggable proteins than the baseline models and current existing methods in terms of the independent test dataset. An online web server was established and made freely available online.
| S-EPMC9421381 | biostudies-literature
Project description:BACKGROUND:Lung cancer is the leading cause of cancer-related deaths in both men and women in the United States, and it has a much lower five-year survival rate than many other cancers. Accurate survival analysis is urgently needed for better disease diagnosis and treatment management. RESULTS:In this work, we propose a survival analysis system that takes advantage of recently emerging deep learning techniques. The proposed system consists of three major components. 1) The first component is an end-to-end cellular feature learning module using a deep neural network with global average pooling. The learned cellular representations encode high-level biologically relevant information without requiring individual cell segmentation, which is aggregated into patient-level feature vectors by using a locality-constrained linear coding (LLC)-based bag of words (BoW) encoding algorithm. 2) The second component is a Cox proportional hazards model with an elastic net penalty for robust feature selection and survival analysis. 3) The third commponent is a biomarker interpretation module that can help localize the image regions that contribute to the survival model's decision. Extensive experiments show that the proposed survival model has excellent predictive power for a public (i.e., The Cancer Genome Atlas) lung cancer dataset in terms of two commonly used metrics: log-rank test (p-value) of the Kaplan-Meier estimate and concordance index (c-index). CONCLUSIONS:In this work, we have proposed a segmentation-free survival analysis system that takes advantage of the recently emerging deep learning framework and well-studied survival analysis methods such as the Cox proportional hazards model. In addition, we provide an approach to visualize the discovered biomarkers, which can serve as concrete evidence supporting the survival model's decision.
| S-EPMC7079513 | biostudies-literature
Project description:Machine learning algorithms are increasingly being utilized to identify brain connectivity biomarkers linked to behavioral and clinical outcomes. However, research often prioritizes prediction accuracy at the expense of biological interpretability, and inconsistent implementation of ML methods may hinder model accuracy. To address this, our paper introduces a network-level enrichment approach, which integrates brain system organization in the context of connectome-wide statistical analysis to reveal network-level links between brain connectivity and behavior. To demonstrate the efficacy of this approach, we used linear support vector regression (LSVR) models to examine the relationship between resting-state functional connectivity networks and chronological age. We compared network-level associations based on raw LSVR weights to those produced from the forward and inverse models. Results indicated that not accounting for shared family variance inflated prediction performance, the k-best feature selection via Pearson correlation reduced accuracy and reliability, and raw LSVR model weights produced network-level associations that deviated from the significant brain systems identified by forward and inverse models. Our findings offer crucial insights for applying machine learning to neuroimaging data, emphasizing the value of network enrichment for biological interpretation.
| S-EPMC11349033 | biostudies-literature