Functional Impact Scores Research Articles

Rare Germline Loss-of-Function Variants in HSP47 ( SERPINH1) Are Associated with an Intermediate Osteogenesis Imperfecta Phenotype Characterized By Atopic Inflammation and Increased Risk of Thrombosis

Background: There has been considerable interest in the collagen-specific chaperone HSP47 ( SERPINH1) as a potential drug target for the treatment of cirrhosis, fibrotic disease, and more recently, thrombosis (Thienel et al., Science 380, 178-187, 2023). While homozygous or compound heterozygous loss of function in SERPINH1 is known to cause a rare form of osteogenesis imperfecta (OI) in humans, little is known about the clinical effects of moderately decreased HSP47 activity. In order to assess the potential safety and efficacy of an antithrombotic strategy based on HSP47 blockade, we evaluated the clinical impacts of heterozygous loss-of-function variants in SERPINH1 in a dataset of over 400,000 individuals. Aims: Determine the clinical effects of SERPINH1 loss of function in a large-scale whole exome sequencing dataset. Methods: The UK Biobank (UKBB) contains paired whole exome sequencing and clinical data on 415,921 subjects in addition to plasma proteomics data for a subset of participants (N= 48,892). We identified all rare (MAF<0.1%)variants in SERPINH1 that were predicted in silico to alter protein activity (functional impact score ≥0.7). Using Firth's logistic regression, we controlled for age, sex, ancestry, and other risk factors and assessed the association between the presence of qualifying SERPINH1 variants and four thrombotic disorders: venous thromboembolism (VTE), non-cardioembolic ischemic stroke (NCEIS), myocardial infarction (MI), and peripheral arterial disease (PAD). Replication was performed in a composite dataset from the NIH All of Us program and the Mass General Brigham (MGB) Biobank (N=150,017). We also evaluated differences in the plasma levels of 1,472 proteins (Olink Explore 1536 proteomics panel) between carriers and non-carriers of qualifying SERPINH1 variants. Results: Rare qualifying variants in SERPINH1 were identified in 382 UKBB participants (100% heterozygous). On average, SERPINH1 variant carriers were of significantly shorter stature (P=3.97 x 10-5) and lower weight (P=0.006) than individuals without such variants, and this effect was more pronounced in subjects with higher functional impact scores ( Figure 1). By contrast, bone mineral density did not differ significantly as measured by calcaneal quantitative ultrasound and femur shaft dual x-ray absorptiometry. The presence of SERPINH1 variants was associated with significantly increased risk of VTE (OR=1.87, 95% CI: 1.21-2.76, P=0.006), MI (OR=1.98, 95% CI: 1.27-2.96, P=0.003), PAD (OR=2.37, 95% CI: 1.29-3.98, P=0.007), and NCEIS (OR=2.67, 95% CI: 1.24-4.98, P=0.015). Restricting the analysis to high-confidence loss-of-function variants (i.e., nonsense, frameshift, and essential splice site mutations) markedly boosted effect size estimates for MI, PAD, and NCEIS. The VTE and MI disease associations replicated in the composite dataset (VTE OR=3.29, 95% CI: 1.46-7.39, P=0.004; MI OR=2.31, 95% CI: 1.05-5.08, P=0.038). Additionally, the association with NCEIS demonstrated a strong trend towards significance in the replication dataset (OR=2.08, 95% CI: 0.90-4.83, P=0.088). Plasma proteome analysis showed a significant increase in several proteins related to asthma, atopic inflammation, and/or eosinophil activation among individuals with qualifying SERPINH1 variants, including RNASE3, DPP10, TSPAN1, CCL20, and IL10RA ( Figure 2). In order to further investigate these differences in circulating protein profile, we assessed the risk of obstructive asthma in a meta-analysis of the UKBB, All of Us, and MGB datasets (total N=565,455) and found that SERPINH1 variant carriers were significantly more likely to have disease (OR=1.32, 95% CI: 1.03-1.69, P=0.026). Conclusions: Loss of function in SERPINH1 is associated with a significantly increased risk of both venous and arterial thrombosis, raising concerns that long-term therapeutic inhibition of HSP47 may not be a safe or effective antithrombotic strategy. Our data indicate that SERPINH1 variant carriers experience higher levels of eosinophil-driven inflammation, which has consistently been linked to cardiovascular disease. Further, heterozygous SERPINH1 variant carriers appear to have an intermediate OI phenotype characterized by shorter stature but without significant differences in bone mineral density compared to non-carriers.

Clinical Consequences of Rare Germline Variation in Genes Associated with Coagulation across 451,958 Individuals

Introduction Despite deep biological insights over the preceding decades, our ability to assess for inherited hypercoagulable states remains limited. In light of these challenges and the increasing relevance of precision medicine and next-generation sequencing (NGS) to clinical practice, the identification of rare germline variants in coagulation-focused NGS panels is increasingly becoming accepted practice. However, the discovery of variants of unknown significance (VUS) remains a common clinical conundrum in molecular testing, particularly because evidence for the thrombosis risk associated with many genes derives from small, uncontrolled studies, case reports, and pedigree analysis. Large-scale evaluation of the clinical consequences associated with loss of function in individual coagulation-related genes would contribute to the clinical assessment of patients presenting with heterozygous VUS in poorly characterized genes. Objectives Determine the risk of four thrombosis phenotypes associated with rare, germline loss-of-function variants in a major clinical coagulation-focused NGS panel. Methods The UK Biobank (UKBB) contains paired clinical and whole exome sequencing data for over 450,000 participants. We identified rare (minor allele frequency <0.1%) germline variants that were predicted in silico to alter protein activity (functional impact score ≥0.8) within the 54 evaluable genes of the Yale NGS coagulation panel. For each gene, parallel multivariable Firth's logistic regression models were generated to assess the association between rare qualifying variants and four thrombotic phenotypes as well as coagulopathic bleeding in a gene-level collapsing analysis (N=270 independent models). We adjusted for age, sex, ancestry, and sequencing batch. Phenotypes of interest included venous thromboembolism (VTE), myocardial infarction (MI), stroke (CVA), peripheral artery disease (PAD), and coagulopathic bleeding. The Bonferroni corrected p-value for each phenotype was 0.001. We validated functional impact score predictions in seven secreted proteins measured in a subset of 47,923 UKBB participants who underwent plasma proteomic assessment (Olink Explore 1536 panel). Results A total of 451,958 individuals in the UKBB with whole exome sequencing data were evaluated. Among the 54 genes in the Yale NGS coagulation panel, there were 18,975 qualifying variants (98.4% heterozygous). Function-altering variants in 4 of 54 genes were associated with a statistically significant increased risk of thrombosis after adjusting for multiple comparisons: PROC (OR = 5.7, 95% CI: 3.2 - 9.6, p = 1.6x10 -7), SERPINC1 (OR = 5.0, 95% CI: 2.7 - 8.7, p = 4.0x10 -6), PROS1 (OR = 3.2, 95% CI: 2.0 - 4.8, p = 6.0x10 -6), and STAB2 (OR = 1.4, 95% CI: 1.2 - 1.7, p = 5.3x10 -4) (Figure 1). Nominal significance was achieved for 7 genes in the VTE phenotype, 1 gene in CVA, 3 genes in MI, 5 genes in PAD, and 0 genes in coagulopathic bleeding. Variants in a total of 37 genes did not appear to be significantly associated with any of the evaluated phenotypes. To assess the reliability of variant functional impact scores derived in silico, plasma levels for 13 of the 53 coagulation genes were analyzed. Of these, 8 were identified as secreted proteins amenable to assessment by plasma levels, including ADAMTS13 ( ADAMTS13), factor IX ( F9), factor VII ( F7), tissue plasminogen activator ( PLAT), urokinase ( PLAU), protein C ( PROC), plasminogen activator inhibitor 1 ( SERPINE1), and von Willebrand factor ( VWF). For these proteins, average plasma levels consistently demonstrated a negative correlation with functional impact score (Pearson r = -0.98, P=0.0009), indicating that our in silico assessments in this context are broadly reliable (Figure 2). Conclusion Within the commonly encountered clinical context of heterozygous loss of function, most genes contained in coagulation-focused NGS panels likely have fairly weak effects on disease risk that are limited to venous thrombosis. However, our findings do not rule out the possibility of more profound effects from homozygous or compound heterozygous loss of function in these same genes. By defining the clinical consequence of heterozygous loss of function in specific coagulation-related genes across a large population, our findings may help inform the interpretation of molecular testing results in which patients are found to have a VUS.

Coagulation Factor XII ( F12) Haploinsufficiency Is Protective Against Venous Thromboembolism in 415,921 Individuals

Background: Coagulation factor XII (FXII) has drawn significant attention as a potential drug target due to the observation that severe congenital deficiency in FXII is not associated with bleeding, while deletion or inhibition of FXII in pre-clinical models consistently protects against thrombosis. However, the contribution of FXII to thrombosis in humans remains controversial, with early reports suggesting that FXII deficiency may even be associated with an increased risk of venous thromboembolism (VTE). Meanwhile, epidemiologic studies of FXII deficiency have yielded conflicting results and are fraught with methodological challenges in part because FXII-deficient patients have no known phenotype and do not typically present to medical attention. To overcome these limitations, we performed the first large-scale analysis of differential thrombotic risk associated with germline loss of function in the gene encoding FXII ( F12) across over 400,000 individuals. Aims: To determine the impact of rare F12 loss-of-function variants on the risk of VTE in a large-scale whole exome sequencing dataset. Methods: The UK Biobank (UKBB) is a population-scale registry dataset containing paired clinical and genomics data for nearly 500,000 individuals. We identified individuals in the UKBB with rare (MAF <1%) variants in F12 that were predicted in silico to result in altered protein function. Qualifying variants included missense variants with an in silico functional impact score of ≥0.9 and high-confidence loss-of-function (HCLOF) variants such as frameshift, nonsense, and essential splice site mutations. We used Firth's logistic regression to assess the association between qualifying F12 variants in a single-gene collapsing analysis and the risk of VTE after adjusting for age, sex, and ancestry as well as several known contributory factors. Differences in the levels of 1,472 plasma proteins between carriers and non-carriers of qualifying F12 variants were evaluated in a subset of 48,892 UKBB participants using the Olink Explore 1536 proteomics platform. Results: Data from 415,921 UKBB participants were included in our analysis. In total, 1,598 individuals carried qualifying variants in F12, of which 99.8% were heterozygous. The presence of rare, function-altering variants in F12 was significantly associated with protection against VTE (OR=0.74, 95% CI: 0.55-0.97, P=0.029) ( Figure 1). This finding remained unchanged when we restricted the analysis to HCLOF variants (OR=0.74, 95% CI: 0.55-0.97, P=0.031). Further, we replicated several known risk factors for VTE, including female sex (P=0.0015), older age (P<1 x 10 -16), higher BMI (P<1 x 10 -16), non-O blood group (P<1 x 10 -16), higher CRP (P<1 x 10 -16), and a history of smoking (P<1 x 10 -16), suggesting that clinical annotation for the dataset is reliable. As expected, the presence of F12 variants was not associated with increased risk of hemorrhage (OR=0.90, P=0.44). Because FXII is also involved in pathways of innate immunity, we interrogated the relationship between carriage of rare F12 variants and sepsis and found no increased risk (OR=0.95, P=0.73). Next, we compared the circulating plasma levels of 1,472 proteins between participants with and without qualifying F12 variants ( Figure 2). We determined that plasma concentrations of three proteins are significantly elevated in participants with a variant F12 allele. Of these, heme binding protein 1 (HEBP1) levels demonstrated the most striking difference (fold-change=7.08, P=1.439 x 10 -43). However, germline loss-of-function variants in the HEBP1 gene are not associated with decreased risk of VTE (OR=1.109, P=0.88), suggesting that HEPB1 levels are regulated by FXII and do not directly influence VTE risk. Conclusions: Operating at the level of germline genetics, our data are the first to demonstrate that FXII loss-of-function is protective against VTE in humans. Because F12 variant carriers in the UKBB dataset predominantly possess only a single defective allele, it is possible that larger effect sizes would be observed among homozygous or compound heterozygous individuals. Additionally, F12 loss-of-function does not appear to increase the risk of hemorrhage or severe infection. Our findings suggest that therapeutic blockade of FXII in humans is likely to be safe and effective.

Systematically testing human HMBS missense variants to reveal mechanism and pathogenic variation

Defects in hydroxymethylbilane synthase (HMBS) can cause acute intermittent porphyria (AIP), an acute neurological disease. Although sequencing-based diagnosis can be definitive, ∼⅓ of clinical HMBS variants are missense variants, and most clinically reported HMBS missense variants are designated as "variants of uncertain significance" (VUSs). Using saturation mutagenesis, en masse selection, and sequencing, we applied a multiplexed validated assay to both the erythroid-specific and ubiquitous isoforms of HMBS, obtaining confident functional impact scores for >84% of all possible amino acid substitutions. The resulting variant effect maps generally agreed with biochemical expectations and provide further evidence that HMBS can function as a monomer. Additionally, the maps implicated specific residues as having roles in active site dynamics, which was further supported by molecular dynamics simulations. Most importantly, these maps can help discriminate pathogenic from benign HMBS variants, proactively providing evidence even for yet-to-be-observed clinical missense variants.

Estimation of Neutral Mutation Rates and Quantification of Somatic Variant Selection Using cancereffectsizeR.

An R package provides streamlined, customizable estimation of underlying nucleotide mutation rates and of the oncogenic and epistatic effects of mutations in cancer cohorts.

Early detection of emerging SARS-CoV-2 variants of interest for experimental evaluation.

Since the beginning of the COVID-19 pandemic, SARS-CoV-2 has demonstrated its ability to rapidly and continuously evolve, leading to the emergence of thousands of different sequence variants, many with distinctive phenotypic properties. Fortunately, the broad application of next generation sequencing (NGS) across the globe has produced a wealth of SARS-CoV-2 genome sequences, offering a comprehensive picture of how this virus is evolving so that accurate diagnostics, reliable therapeutics, and prophylactic vaccines against COVID-19 can be developed and maintained. The millions of SARS-CoV-2 sequences deposited into genomic sequencing databases, including GenBank, BV-BRC, and GISAID, are annotated with the dates and geographic locations of sample collection, and can be aligned to and compared with the Wuhan-Hu-1 reference genome to extract their constellation of nucleotide and amino acid substitutions. By aggregating these data into concise datasets, the spread of variants through space and time can be assessed. Variant tracking efforts have initially focused on the Spike protein due to its critical role in viral tropism and antibody neutralization. To identify emerging variants of concern as early as possible, we developed a computational pipeline to process the genomic data and assign risk scores based on both epidemiological and functional parameters. Epidemiological dynamics are used to identify variants exhibiting substantial growth over time and spread across geographical regions. Experimental data that quantify Spike protein regions targeted by adaptive immunity and critical for other virus characteristics are used to predict variants with consequential immunogenic and pathogenic impacts. The growth assessment and functional impact scores are combined to produce a Composite Score for any set of Spike substitutions detected. With this systematic method to routinely score and rank emerging variants, we have established an approach to identify threatening variants early and prioritize them for experimental evaluation.

OMEN: network-based driver gene identification using mutual exclusivity.

Network-based driver identification methods that can exploit mutual exclusivity typically fail to detect rare drivers because of their statistical rigor. Propagation-based methods in contrast allow recovering rare driver genes, but the interplay between network topology and high-scoring nodes often results in spurious predictions. The specificity of driver gene detection can be improved by taking into account both gene-specific and gene-set properties. Combining these requires a formalism that can adjust gene-set properties depending on the exact network context within which a gene is analyzed. We developed OMEN: a logic programming framework based on random walk semantics. OMEN presents a number of novel concepts. In particular, its design is unique in that it presents an effective approach to combine both gene-specific driver properties and gene-set properties, and includes a novel method to avoid restrictive, a priori filtering of genes by exploiting the gene-set property of mutual exclusivity, expressed in terms of the functional impact scores of mutations, rather than in terms of simple binary mutation calls. Applying OMEN to a benchmark dataset derived from TCGA illustrates how OMEN is able to robustly identify driver genes and modules of driver genes as proxies of driver pathways. The source code is freely available for download at www.github.com/DriesVanDaele/OMEN. The dataset is archived at https://doi.org/10.5281/zenodo.6419097 and the code at https://doi.org/10.5281/zenodo.6419764. Supplementary data are available at Bioinformatics online.

FI-Net: Identification of Cancer Driver Genes by Using Functional Impact Prediction Neural Network.

Identification of driver genes, whose mutations cause the development of tumors, is crucial for the improvement of cancer research and precision medicine. To overcome the problem that the traditional frequency-based methods cannot detect lowly recurrently mutated driver genes, researchers have focused on the functional impact of gene mutations and proposed the function-based methods. However, most of the function-based methods estimate the distribution of the null model through the non-parametric method, which is sensitive to sample size. Besides, such methods could probably lead to underselection or overselection results. In this study, we proposed a method to identify driver genes by using functional impact prediction neural network (FI-net). An artificial neural network as a parametric model was constructed to estimate the functional impact scores for genes, in which multi-omics features were used as the multivariate inputs. Then the estimation of the background distribution and the identification of driver genes were conducted in each cluster obtained by the hierarchical clustering algorithm. We applied FI-net and other 22 state-of-the-art methods to 31 datasets from The Cancer Genome Atlas project. According to the comprehensive evaluation criterion, FI-net was powerful among various datasets and outperformed the other methods in terms of the overlap fraction with Cancer Gene Census and Network of Cancer Genes database, and the consensus in predictions among methods. Furthermore, the results illustrated that FI-net can identify known and potential novel driver genes.

FrDriver: A Functional Region Driver Identification for Protein Sequence.

Identifying cancer drivers is a crucial challenge to explain the underlying mechanisms of cancer development. There are many methods to identify cancer drivers based on the single mutation site or the entire gene. But they ignore a large number of functional elements with medium in size. It is hypothesized that mutations occurring in different regions of the protein sequence have different effects on the progression of cancer. Here, we develop a novel functional region driver(frDriver) identification method based on Bayesian probability and multiple linear regression models to identify protein regions that can regulate gene expression levels and have high functional impact potential. Combining gene expression data and somatic mutation data, with functional impact scores(SIFT, PROVEAN) as a priori knowledge, we identified cancer driver regions that are most accurate in predicting gene expression levels. We evaluated the performance of frDriver on the BRCA and GBM datasets from TCGA. The results showed that frDriver identified known cancer drivers and outperformed the other three state-of-the-art methods(eDriver, ActiveDriver and OncodriveCLUST). In addition, we performed KEGG pathway and GO term enrichment analysis, and the results indicated that the cancer drivers predicted by frDriver were related to processes such as cancer formation and gene regulation.

Functional oral health-related quality of life impact: A systematic review in populations with tooth loss.

Patients perceive the impact of oral disorder in four major areas, the dimensions of oral health-related quality life (OHRQoL) Oral Function, Orofacial Pain, Orofacial Appearance, and Psychosocial Impact. The functional aspect is essential given the need of chewing, biting, speech and swallowing. The objective of this study was to identify OHRQoL information for dental subjects with functional oral health problems. In a systematic review, distinct and clinically relevant groups of dental subject samples, in this study called "population groups," with such functional OHRQoL information based on the Oral Health Impact Profile (OHIP) were identified (PROSPERO registration: CRD42017064033). The search strategy was "Oral Health Impact Profile" or OHIP. Searches were conducted in the PubMed interface of the Medline database, EMBASE, Cochrane, CINAHL and PsyINFO on 8 June 2017 and updated on 14 January 2019. Published OHIP domain data of different versions were recalculated into OHIP-14`s Physical Disability domain score, characterising the subject's Oral Function impact. 3,653 potentially abstracts were screened. We identified 78 publications reporting dimensional information on 154 subject samples with 52 populations. A typical mean functional impact for partially dentate subjects was 1.6 units on a 0 to 8 unit metric, while for edentate subjects, the mean functional impact was 2.6 units. The functional impact score ranged from 0 to 7.9 units with 50% of the patient samples located between 0.8 and 2.6 units. For the first time, we provide normalised and therefore comparable metric information about the functional OHRQoL impact for a substantial number of functional oral conditions.

Combined burden and functional impact tests for cancer driver discovery using DriverPower

The discovery of driver mutations is one of the key motivations for cancer genome sequencing. Here, as part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium, which aggregated whole genome sequencing data from 2658 cancers across 38 tumour types, we describe DriverPower, a software package that uses mutational burden and functional impact evidence to identify driver mutations in coding and non-coding sites within cancer whole genomes. Using a total of 1373 genomic features derived from public sources, DriverPower’s background mutation model explains up to 93% of the regional variance in the mutation rate across multiple tumour types. By incorporating functional impact scores, we are able to further increase the accuracy of driver discovery. Testing across a collection of 2583 cancer genomes from the PCAWG project, DriverPower identifies 217 coding and 95 non-coding driver candidates. Comparing to six published methods used by the PCAWG Drivers and Functional Interpretation Working Group, DriverPower has the highest F1 score for both coding and non-coding driver discovery. This demonstrates that DriverPower is an effective framework for computational driver discovery.

Differential Allele-Specific Expression Uncovers Breast Cancer Genes Dysregulated by Cis Noncoding Mutations.

Identifying cancer-relevant mutations in noncoding regions is challenging due to the large numbers of such mutations, their low levels of recurrence, and difficulties in interpreting their functional impact. To uncover genes that are dysregulated due to somatic mutations in cis, we build upon the concept of differential allele-specific expression (ASE) and introduce methods to identify genes within an individual's cancer whose ASE differs from what is found in matched normal tissue. When applied to breast cancer tumor samples, our methods detect the known allele-specific effects of copy number variation and nonsense-mediated decay. Further, genes that are found to recurrently exhibit differential ASE across samples are cancer relevant. Genes with cis mutations are enriched for differential ASE, and we find 147 potentially functional noncoding mutations cis to genes that exhibit significant differential ASE. We conclude that differential ASE is a promising means for discovering gene dysregulation due to cis noncoding mutations.

An integrative machine learning approach for prediction of toxicity-related drug safety.

Recent trends in drug development have been marked by diminishing returns caused by the escalating costs and falling rates of new drug approval. Unacceptable drug toxicity is a substantial cause of drug failure during clinical trials and the leading cause of drug withdraws after release to the market. Computational methods capable of predicting these failures can reduce the waste of resources and time devoted to the investigation of compounds that ultimately fail. We propose an original machine learning method that leverages identity of drug targets and off-targets, functional impact score computed from Gene Ontology annotations, and biological network data to predict drug toxicity. We demonstrate that our method (TargeTox) can distinguish potentially idiosyncratically toxic drugs from safe drugs and is also suitable for speculative evaluation of different target sets to support the design of optimal low-toxicity combinations.

Abstract 1279: Passenger mutation landscape in cancer genomes

Abstract Cancer progression is an evolutionary process during which thousands of somatic variants are accumulated within an individual. In the classic view of cancer progression, a handful of driver variants are thought to give a positive selection advantage to the cancer cell. The remaining variants, termed passengers, represent the overwhelming majority of the variants in cancer genomes, and their functional consequences are poorly understood. Furthermore, the bulk of these passengers fall within noncoding regions of the genome, making these the main product of whole-genome sequencing of tumors. In this work, we explore the functional landscape of passenger variants in various cancer cohorts by leveraging extensive pan-cancer variant calls from ~2500 uniformly processed whole cancer genomes. More specifically, we integrate their annotations and predicted functional impact scores to quantify the overall burdening of various elements in cancer genomes. We show that disruption of genetic regulatory elements in the noncoding genome correlates with altered gene expression. Furthermore, we also show how overall functional burdening of various genomic elements correlate with patient survival time, and tumor clonality. Finally, we observe statistical signals consistent with the notion that aggregated subsets of passenger variants - particularly those we predict to be functionally impactful- might confer weak selective effects. Citation Format: Sushant Kumar, Jonathan Warrel, Patrick Mcgillivray, William Meyerson, Shantao Li, Leonidas Salichos, Arif Harmanci, Alexander Fundichely, Calvin Chan, Carl Herrmann, Morten Nielsen, Lucas Lochovsky, Yan Zhang, Xiaotong Li, Ekta Khurana, Gad Getz, Mark Gerstein. Passenger mutation landscape in cancer genomes [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 1279.

Cancer driver mutation prediction through Bayesian integration of multi-omic data.

Identification of cancer driver mutations is critical for advancing cancer research and personalized medicine. Due to inter-tumor genetic heterogeneity, many driver mutations occur at low frequencies, which make it challenging to distinguish them from passenger mutations. Here, we show that a novel Bayesian hierarchical modeling approach, named rDriver can achieve enhanced prediction accuracy by identifying mutations that not only have high functional impact scores but also are associated with systemic variation in gene expression levels. In examining 3,080 tumor samples from 8 cancer types in The Cancer Genome Atlas, rDriver predicted 1,389 driver mutations. Compared with existing tools, rDriver identified more low frequency mutations associated with lineage specific functional properties, timing of occurrence and patient survival. Evaluation of rDriver predictions using engineered cell-line models resulted in a positive predictive value of 0.94 in PIK3CA genes. Our study highlights the importance of integrating multi-omic data in predicting cancer driver mutations and provides a statistically rigorous solution for cancer target discovery and development.

Significance of the TMPRSS2:ERG gene fusion in prostate cancer.

The transmembrane protease serine 2:v-ets erythroblastosis virus E26 oncogene homolog (TMPRSS2:ERG) gene fusion is common in prostate cancer, while its functional role is not fully understood. The present study aimed to investigate the significance of the TMPRSS2:ERG gene fusion in human prostate cancers using bioinformatics tools. Comprehensive alteration analysis of TMPRSS2 and ERG in 148 different human cancer studies was performed by cBioPortal, and the mRNA expression level of the ERG gene was evaluated using Oncomine analysis. Furthermore, lentiviral short hairpin (sh)RNA-mediated knockdown of TMPRSS2:ERG was performed to study the impact of ERG silencing on cell proliferation and cell cycle distribution in prostate cancer cells. The results demonstrated that the TMPRSS2 and ERG genes were mostly altered in prostate cancer, and the most frequent alteration was gene fusion. Oncomine analysis demonstrated that the ERG gene was significantly upregulated in prostate clinical samples compared with the normal prostate gland in four independent datasets, and a positive association was observed between potassium inwardly-rectifying channel subfamily J member 15, down syndrome critical region gene 4, potassium inwardly-rectifying channel subfamily J member 6 and ERG gene expression. There were 272 mutations of the ERG gene identified in the cBioPortal database; among the mutations, 2 missense mutations (R367C and P401H) were regarded as functional mutations (functional impact score >1.938). Furthermore, the present study successfully knocked down ERG gene expression through a lentiviral-mediated gene silencing approach in VCaP prostate cancer cells. The ERG mRNA and protein expression levels were both suppressed significantly, and a cell-cycle arrest at G0/G1 phase was observed after ERG gene silencing. In conclusion, these bioinformatics analyses provide novel insights for TMPRSS2:ERG fusion gene study in prostate cancer. Target inhibition of ERG expression could significantly cause cell growth arrest in prostate cancer cells, which could be a potentially valuable target for prostate cancer treatment. However, the precise mechanism of these results remains unclear; therefore, further studies are required.

Abstract A1-33: Knowledge prioritization of cancer genomes using oncoscores and functional impact scores to support biomarker discovery and clinical decision making

Abstract Introduction: To accelerate the translation of genome information in important biological and clinical insights requires systematic integration and analysis of functional genomic and proteomic information. In this way molecular landscapes of cancer patients can be characterized efficiently to impact their care. To fully rationalize the potential treatment options for patients based on mechanistic considerations, we must understand two key parameters: (a) how important is each protein/gene to the known mechanisms underlying specific cancer types and (b) what are the functional implications of variants at the level of protein and system function. Of the many modules in our pipeline, here we present two of the components we have developed to address these questions. Methods: Based on expert-curated data, combined with bioinformatics and text-data mining methods we have modeled (a) the relative biological importance of each gene/protein to specific cancer types, with a measure we call ‘oncoscore’ (b) the functional effects of a variant with a process we call ‘functional impact score’. First, the oncoscore method relies on multidimensional data types that summarize real-time evidence regarding clinical and molecular importance with respect to specific cancer indications. Such features include, among others, gene/protein pathway inclusion facts, drug-targetness, disease association, interaction neighborhood, as well as indication-specific protein and druggability attributes. Applying these parameters across individual cancer indications provides us with a prioritization of the functionally most important genes associated with each cancer type. Secondly, to understand the impact of aberrations we contextualized structural, functional, drug response and safety information to provide a novel approach for the prediction of functionally important aberrations. Combining the two scores permits prioritization of the functionally most important genes and aberrations in any patient tumor. Interestingly, the strategy can also be applied in absence of DNA sequence information, where the oncoscore method alone can be used to prioritize potential drug targets, again based on levels of real-time evidence. Results: We present a subset of our results in the context of 25 different cancer conditions and demonstrate how the two scores can help prioritize the most important clinico-molecular players of a disease, decipher the most important aberrations found in patients' molecular profiles, and respectively combine this information to prioritize treatments for each indication/patient. Conclusion: While our database contains curated information about the relationship between a gene/protein mutation and drug response within specific cancer types, we have devised two additional mechanisms to expand the clinical actionability of this information. The oncoscore and functional-impact scores provide an additional modus to decipher clinically actionable information from a patient tumor, especially when no known biomarkers are detected in the patient's profile. These methods are also particularly applicable to the identification of novel treatment biomarkers. Another advantage is that they can be used to prioritize patient treatments in the absence of sequence information, a feature that can be helpful when it comes to non-resectable disease in rare cancers. In summary, our cancer-specific integration of biological and clinical knowledge allows us to predict potentially actionable mutations in patient tumors. This is an important extension to the identification of previously known predictive biomarkers and lends itself to translational level clinical applications including biomarker discovery, drug repositioning and clinical trial prioritization for ex-guideline patients. Citation Format: Theodoros G. Soldatos, Sasha Badbanchi, Sonia Vivas, Alexander Zien, Francesca Diella, Markus Hartenfeller, Alexander Picker, Martin A. Stein, David B. Jackson. Knowledge prioritization of cancer genomes using oncoscores and functional impact scores to support biomarker discovery and clinical decision making. [abstract]. In: Proceedings of the AACR Special Conference on Translation of the Cancer Genome; Feb 7-9, 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 1):Abstract nr A1-33.

Abstract 4263: Identifying somatic mutation hotspots across protein family alignments

Abstract In cancer, somatic driver mutations often target “hotspots” of paralogous residues across evolutionarily related members of a single protein family. These hotspots usually confer a dominant oncogenic function, such as constitutive catalytic activity or binding. These hotspots generally represent the most “druggable” genetic alterations. Examples of protein families targeted in such a manner include the Ras and receptor Tyrosine Kinase family. Protein family hotspots are not directly assessed by standard computational approaches for statistically nominating genes under positive somatic selection, such as MutSig, InVex, or MuSic. These methods only assess single genes for statistical enrichment of total mutation burden or the presence of mutation hotspots. As a result, such approaches are only powered to detect protein family hotspots in cases where each (or at least one) protein family member is very frequently mutated (e.g. KRAS, NRAS, HRAS). They are not powered to detect hotspots in highly functionally redundant families, of which each member may be only mutated in <<1% of a tumor type. To directly identify such protein-family hotspots, we have developed a computational framework, MutPfam. This framework superimposes mutation data onto Pfam (http://pfam.sanger.ac.uk/) (protein family and subfamily multiple sequence alignments and identifies statistically significant hotspots. We demonstate our method on protein families involved in the catalysis of small-molecular catabolism using the TCGA Pan-Cancer dataset (https://tcga-data.nci.nih.gov/) comprising somatic mutation calls for 4608 whole-exome sequenced patients and 21 tumor types. As we demonstrate, our framework enables the de novo discovery of mutation hotspots across protein families. Though developed for somatic mutation analysis, this framework could also be applied to the analysis of inherited mutations. Computational issues meriting future investigation include (1) alternate definitions of protein alignments and sub-families (e.g. involving three dimensional structure, biophysical constraints) (2) improved statistical modeling of the local background mutation rate (incorporating local genomic covariates that mediate heterogeneity) (3) incorporation of variant functional impact scores for missense variants. Citation Format: Marcin Imielinski, Charles Du, Matthew Meyerson. Identifying somatic mutation hotspots across protein family alignments. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4263. doi:10.1158/1538-7445.AM2014-4263

Abstract A22: Identifying somatic mutation hotspots across protein family alignments.

Abstract In cancer, somatic driver mutations often target “hotspots” of paralogous residues across evolutionarily related members of a single protein family. Mutation hotspots usually confer a dominant oncogenic function and are among the most druggable cancer sequence alterations (e.g. EGFR L858R, BRAF V600E). Examples of protein families targeted in such a manner include the Ras and receptor tyrosine kinase families (1). Protein family hotspots are not directly assessed by standard computational approaches for statistically nominating genes under positive somatic selection in cancer, such as MutSig (2), InVex (3), or MuSic (4). These methods only assess single genes for statistical enrichment of total mutation burden or the presence of mutation hotspots, and thus are only powered to detect protein family hotspots in cases where each (or at least one) protein family member is very frequently mutated (e.g. KRAS, NRAS, HRAS). In particular, they are not powered to detect hotspots in highly functionally redundant families, of which each member may be only mutated in <<1% of cases. To directly identify protein-family hotspots, we have developed a computational framework, MutPfam. This framework superimposes mutation data onto Pfam (http://pfam.sanger.ac.uk/) protein family and subfamily multiple sequence alignments and identifies statistically significant hotspots. We demonstrate our method on the Ras Pfam family (PF00071.17) using TCGA Pan-Cancer dataset (https://tcga-data.nci.nih.gov/) comprising somatic mutation calls for 4608 whole-exome sequenced patients and 21 tumor types. Using phylogeny analysis of this 134-member protein family, we analyzed 110 subfamilies for the presence of significantly mutated hotspots. In addition to canonical KRAS, NRAS, and HRAS hotspots in the switch I (e.g. KRAS G12V, KRAS G13D) and switch II (e.g. HRAS Q61R) domains, we identified several other hotspots achieving significance following Bonferonni correction. Among our top hits was a recently identified melanoma hotspot in RAC1 (3), previously uncharacterized paralogues of a known hotspot in RHOA (RHOA E40Q, RALA D49G, RND1 E48K), and a novel recurrently mutated position in the RAB subfamily (e.g. RAB10 R28C, RAB11B R30H, RAB17 R38W). As we demonstrate, our framework enables the de novo discovery of mutation hotspots across protein families, including those that are rarely mutated in any single family member. Computational issues meriting future investigation include (1) alternate definitions of protein alignments and sub-families (e.g. involving three dimensional structure, biophysical constraints) (2) improved statistical modeling of the local background mutation rate (3) incorporation of variant functional impact scores for missense variants. Citation Information: Mol Cancer Ther 2013;12(11 Suppl):A22. Citation Format: Marcin Imielinski, Charles Du, Matthew Meyerson. Identifying somatic mutation hotspots across protein family alignments. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2013 Oct 19-23; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2013;12(11 Suppl):Abstract nr A22.

Revealing selection in cancer using the predicted functional impact of cancer mutations. Application to nomination of cancer drivers

Every malignant tumor has a unique spectrum of genomic alterations including numerous protein mutations. There are also hundreds of personal germline variants to be taken into account. The combinatorial diversity of potential cancer-driving events limits the applicability of statistical methods to determine tumor-specific "driver" alterations among an overwhelming majority of "passengers". An alternative approach to determining driver mutations is to assess the functional impact of mutations in a given tumor and predict drivers based on a numerical value of the mutation impact in a particular context of genomic alterations.Recently, we introduced a functional impact score, which assesses the mutation impact by the value of entropic disordering of the evolutionary conservation patterns in proteins. The functional impact score separates disease-associated variants from benign polymorphisms with an accuracy of ~80%. Can the score be used to identify functionally important non-recurrent cancer-driver mutations? Assuming that cancer-drivers are positively selected in tumor evolution, we investigated how the functional impact score correlates with key features of natural selection in cancer, such as the non-uniformity of distribution of mutations, the frequency of affected tumor suppressors and oncogenes, the frequency of concurrent alterations in regions of heterozygous deletions and copy gain; as a control, we used presumably non-selected silent mutations. Using mutations of six cancers studied in TCGA projects, we found that predicted high-scoring functional mutations as well as truncating mutations tend to be evolutionarily selected as compared to low-scoring and silent mutations. This result justifies prediction of mutations-drivers using a shorter list of predicted high-scoring functional mutations, rather than the "long tail" of all mutations.