Multiomic analysis identifies a high-risk signature that predicts early clinical failure in DLBCL.

Recent genetic and molecular classification of DLBCL has advanced our knowledge of disease biology, yet were not designed to predict early events and guide anticipatory selection of novel therapies. To address this unmet need, we used an integrative multiomic approach to identify a signature at diagnosis that will identify DLBCL at high risk of early clinical failure. Tumor biopsies from 444 newly diagnosed DLBCL were analyzed by WES and RNAseq. A combination of weighted gene correlation network analysis and differential gene expression analysis was used to identify a signature associated with high risk of early clinical failure independent of IPI and COO. Further analysis revealed the signature was associated with metabolic reprogramming and identified cases with a depleted immune microenvironment. Finally, WES data was integrated into the signature and we found that inclusion of ARID1A mutations resulted in identification of 45% of cases with an early clinical failure which was validated in external DLBCL cohorts. This novel and integrative approach is the first to identify a signature at diagnosis, in a real-world cohort of DLBCL, that identifies patients at high risk for early clinical failure and may have significant implications for design of therapeutic options.

Comprehensive profiling of cancer neoantigens from aberrant RNA splicing.

BACKGROUND: Cancer neoantigens arise from protein-altering somatic mutations in tumor and rank among the most promising next-generation immuno-oncology agents when used in combination with immune checkpoint inhibitors. We previously developed a computational framework, REAL-neo, for identification, quality control, and prioritization of both class-I and class-II human leucocyte antigen (HLA)-presented neoantigens resulting from somatic single-nucleotide mutations, small insertions and deletions, and gene fusions. In this study, we developed a new module, SPLICE-neo, to identify neoantigens from aberrant RNA transcripts from two distinct sources: (1) DNA mutations within splice sites and (2) de novo RNA aberrant splicings. METHODS: First, SPLICE-neo was used to profile all DNA splice-site mutations in 11,892 tumors from The Cancer Genome Atlas (TCGA) and identified 11 profiles of splicing donor or acceptor site gains or losses. Transcript isoforms resulting from the top seven most frequent profiles were computed using novel logic models. Second, SPLICE-neo identified de novo RNA splicing events using RNA sequencing reads mapped to novel exon junctions from either single, double, or multiple exon-skipping events. The aberrant transcripts from both sources were then ranked based on isoform expression levels and z-scores assuming that individual aberrant splicing events are rare. Finally, top-ranked novel isoforms were translated into protein, and the resulting neoepitopes were evaluated for neoantigen potential using REAL-neo. The top splicing neoantigen candidates binding to HLA-A*02:01 were validated using in vitro T2 binding assays. RESULTS: We identified abundant splicing neoantigens in four representative TCGA cancers: BRCA, LUAD, LUSC, and LIHC. In addition to their substantial contribution to neoantigen load, several splicing neoantigens were potent tumor antigens with stronger bindings to HLA compared with the positive control of antigens from influenza virus. CONCLUSIONS: SPLICE-neo is the first tool to comprehensively identify and prioritize splicing neoantigens from both DNA splice-site mutations and de novo RNA aberrant splicings. There are two major advances of SPLICE-neo. First, we developed novel logic models that assemble and prioritize full-length aberrant transcripts from DNA splice-site mutations. Second, SPLICE-neo can identify exon-skipping events involving more than two exons, which account for a quarter to one-third of all skipping events.

Cell-specific crosstalk proteomics reveals cathepsin B signaling as a driver of glioblastoma malignancy near the subventricular zone.

Glioblastoma (GBM) is the most prevalent and aggressive malignant primary brain tumor. GBM proximal to the lateral ventricles (LVs) is more aggressive, potentially due to subventricular zone (SVZ) contact. Despite this, crosstalk between GBM and neural stem/progenitor cells (NSC/NPCs) is not well understood. Using cell-specific proteomics, we show that LV-proximal GBM prevents neuronal maturation of NSCs through induction of senescence. Additionally, GBM brain tumor initiating cells (BTICs) increase expression of CTSB upon interaction with NPCs. Lentiviral knockdown and recombinant protein experiments reveal both cell-intrinsic and soluble CTSB promote malignancy-associated phenotypes in BTICs. Soluble CTSB stalls neuronal maturation in NPCs while promoting senescence, providing a link between LV-tumor proximity and neurogenesis disruption. Finally, we show LV-proximal CTSB upregulation in patients, showing the relevance of this crosstalk in human GBM biology. These results demonstrate the value of proteomic analysis in tumor microenvironment research and provide direction for new therapeutic strategies in GBM. Highlights: Periventricular GBM is more malignant and disrupts neurogenesis in a rodent model.Cell-specific proteomics elucidates tumor-promoting crosstalk between GBM and NPCs.NPCs induce upregulated CTSB expression in GBM, promoting tumor progression.GBM stalls neurogenesis and promotes NPC senescence via CTSB.

Multiomic Analysis Identifies a High-Risk Metabolic and TME Depleted Signature that Predicts Early Clinical Failure in DLBCL.

PURPOSE: 60-70% of newly diagnosed diffuse large B-cell lymphoma (DLBCL) patients avoid events within 24 months of diagnosis (EFS24) and the remainder have poor outcomes. Recent genetic and molecular classification of DLBCL has advanced our knowledge of disease biology, yet were not designed to predict early events and guide anticipatory selection of novel therapies. To address this unmet need, we used an integrative multiomic approach to identify a signature at diagnosis that will identify DLBCL at high risk of early clinical failure. PATIENTS AND METHODS: Tumor biopsies from 444 newly diagnosed DLBCL were analyzed by WES and RNAseq. A combination of weighted gene correlation network analysis and differential gene expression analysis followed by integration with clinical and genomic data was used to identify a multiomic signature associated with high risk of early clinical failure. RESULTS: Current DLBCL classifiers are unable to discriminate cases who fail EFS24. We identified a high risk RNA signature that had a hazard ratio (HR, 18.46 [95% CI 6.51-52.31] P < .001) in a univariate model, which did not attenuate after adjustment for age, IPI and COO (HR, 20.8 [95% CI, 7.14-61.09] P < .001). Further analysis revealed the signature was associated with metabolic reprogramming and a depleted immune microenvironment. Finally, WES data was integrated into the signature and we found that inclusion of ARID1A mutations resulted in identification of 45% of cases with an early clinical failure which was validated in external DLBCL cohorts. CONCLUSION: This novel and integrative approach is the first to identify a signature at diagnosis that will identify DLBCL at high risk for early clinical failure and may have significant implications for design of therapeutic options.

AT-101 Enhances the Antitumor Activity of Lenalidomide in Patients with Multiple Myeloma.

Bcl-2 and Mcl-1 proteins play a role in multiple myeloma (MM) cell survival, for which targeted inhibitors are being developed. AT-101 is an oral drug, which disrupts Bcl-2 and Mcl-1 function, impedes mitochondrial bioenergetic processes and induces apoptosis in MM cells. When combined with lenalidomide and dexamethasone (Rd), AT-101 significantly reduced tumor burden in an in vivo xenograft model of MM. These data provided rationale for a phase I/II study to establish the effective dose of AT-101 in combination with Rd (ARd regimen) in relapsed/refractory MM. A total of 10 patients were enrolled, most with high-risk cytogenetics (80%) and prior stem cell transplant (70%). Three patients were lenalidomide-refractory, 2 were bortezomib-refractory and 3 were daratumumab-refractory. The ARd combination was well tolerated with most common grade 3/4 adverse events being cytopenia's. The overall response rate was 40% and clinical benefit rate was 90%. The median progression free survival was 14.9 months (95% CI 7.1-NE). Patients responsive to ARd showed a decrease in Bcl-2:Bim or Mcl-1:Noxa protein complexes, increased CD8+ T and NK cells and depletion of T and B-regulatory cells. The ARd regimen demonstrated an acceptable safety profile and promising efficacy in patients with relapsed/refractory MM prompting further investigation in additional patients.

Elevating microglia TREM2 reduces amyloid seeding and suppresses disease-associated microglia.

TREM2 is exclusively expressed by microglia in the brain and is strongly linked to the risk for Alzheimer's disease (AD). As microglial responses modulated by TREM2 are central to AD pathogenesis, enhancing TREM2 signaling has been explored as an AD therapeutic strategy. However, the effective therapeutic window targeting TREM2 is unclear. Here, by using microglia-specific inducible mouse models overexpressing human wild-type TREM2 (TREM2-WT) or R47H risk variant (TREM2-R47H), we show that TREM2-WT expression reduces amyloid deposition and neuritic dystrophy only during the early amyloid seeding stage, whereas TREM2-R47H exacerbates amyloid burden during the middle amyloid rapid growth stage. Single-cell RNA sequencing reveals suppressed disease-associated microglia (DAM) signature and reduced DAM population upon TREM2-WT expression in the early stage, whereas upregulated antigen presentation pathway is detected with TREM2-R47H expression in the middle stage. Together, our findings highlight the dynamic effects of TREM2 in modulating AD pathogenesis and emphasize the beneficial effect of enhancing TREM2 function in the early stage of AD development.

Assessment and Optimization of Explainable Machine Learning Models Applied to Transcriptomic Data.

Explainable artificial intelligence aims to interpret how machine learning models make decisions, and many model explainers have been developed in the computer vision field. However, understanding of the applicability of these model explainers to biological data is still lacking. In this study, we comprehensively evaluated multiple explainers by interpreting pre-trained models for predicting tissue types from transcriptomic data and by identifying the top contributing genes from each sample with the greatest impacts on model prediction. To improve the reproducibility and interpretability of results generated by model explainers, we proposed a series of optimization strategies for each explainer on two different model architectures of multilayer perceptron (MLP) and convolutional neural network (CNN). We observed three groups of explainer and model architecture combinations with high reproducibility. Group II, which contains three model explainers on aggregated MLP models, identified top contributing genes in different tissues that exhibited tissue-specific manifestation and were potential cancer biomarkers. In summary, our work provides novel insights and guidance for exploring biological mechanisms using explainable machine learning models.

Lower Exome Sequencing Coverage of Ancestrally African Patients in The Cancer Genome Atlas.

BACKGROUND: In the United States, cancer disproportionately impacts Black and African American individuals. Identifying genetic factors underlying cancer disparities has been an important research focus and requires data that are equitable in both quantity and quality across racial groups. It is widely recognized that DNA databases quantitatively underrepresent minorities. However, the differences in data quality between racial groups have not been well studied. METHODS: We compared the qualities of germline and tumor exomes between ancestrally African and European patients in The Cancer Genome Atlas of 7 cancers with at least 50 self-reported Black patients in the context of sequencing depth, tumor purity, and qualities of germline variants and somatic mutations. RESULTS: Germline and tumor exomes from ancestrally African patients were sequenced at statistically significantly lower depth in 6 out of the 7 cancers. For 3 cancers, most ancestrally European exomes were sequenced in early sample batches at higher depth, whereas ancestrally African exomes were concentrated in later batches and sequenced at much lower depth. For the other 3 cancers, the reasons of lower sequencing coverage of ancestrally African exomes remain unknown. Furthermore, even when the sequencing depths were comparable, African exomes had disproportionally higher percentages of positions with insufficient coverage, likely because of the known European bias in the human reference genome that impacted exome capture kit design. CONCLUSIONS: Overall and positional lower sequencing depths of ancestrally African exomes in The Cancer Genome Atlas led to underdetection and lower quality of variants, highlighting the need to consider epidemiological factors for future genomics studies.

Shared brain transcriptomic signature in TDP-43 type A FTLD patients with or without GRN mutations.

Frontotemporal lobar degeneration with TDP-43 inclusions (FTLD-TDP) is a complex heterogeneous neurodegenerative disorder for which mechanisms are poorly understood. To explore transcriptional changes underlying FTLD-TDP, we performed RNA-sequencing on 66 genetically unexplained FTLD-TDP patients, 24 FTLD-TDP patients with GRN mutations and 24 control participants. Using principal component analysis, hierarchical clustering, differential expression and coexpression network analyses, we showed that GRN mutation carriers and FTLD-TDP-A patients without a known mutation shared a common transcriptional signature that is independent of GRN loss-of-function. After combining both groups, differential expression as compared to the control group and coexpression analyses revealed alteration of processes related to immune response, synaptic transmission, RNA metabolism, angiogenesis and vesicle-mediated transport. Deconvolution of the data highlighted strong cellular alterations that were similar in FTLD-TDP-A and GRN mutation carriers with NSF as a potentially important player in both groups. We propose several potentially druggable pathways such as the GABAergic, GDNF and sphingolipid pathways. Our findings underline new disease mechanisms and strongly suggest that affected pathways in GRN mutation carriers extend beyond GRN and contribute to genetically unexplained forms of FTLD-TDP-A.

Role of PLEXIND1/TGFß Signaling Axis in Pancreatic Ductal Adenocarcinoma Progression Correlates with the Mutational Status of KRAS.

PLEXIND1 is upregulated in several cancers, including pancreatic ductal adenocarcinoma (PDAC). It is an established mediator of semaphorin signaling, and neuropilins are its known coreceptors. Herein, we report data to support the proposal that PLEXIND1 acts as a transforming growth factor beta (TGFß) coreceptor, modulating cell growth through SMAD3 signaling. Our findings demonstrate that PLEXIND1 plays a pro-tumorigenic role in PDAC cells with oncogenic KRAS (KRASmut). We show in KRASmut PDAC cell lines (PANC-1, AsPC-1,4535) PLEXIND1 downregulation results in decreased cell viability (in vitro) and reduced tumor growth (in vivo). Conversely, PLEXIND1 acts as a tumor suppressor in the PDAC cell line (BxPC-3) with wild-type KRAS (KRASwt), as its reduced expression results in higher cell viability (in-vitro) and tumor growth (in vivo). Additionally, we demonstrate that PLEXIND1-mediated interactions can be selectively disrupted using a peptide based on its C-terminal sequence (a PDZ domain-binding motif), an outcome that may possess significant therapeutic implications. To our knowledge, this is the first report showing that (1) PLEXIND1 acts as a TGFß coreceptor and mediates SMAD3 signaling, and (2) differential roles of PLEXIND1 in PDAC cell lines correlate with KRASmut and KRASwt status.

The Prognostic Role of MYC Structural Variants Identified by NGS and FISH in Multiple Myeloma.

PURPOSE: Structural variants (SV) of the MYC gene region are common in multiple myeloma and influence disease progression. However, the prognostic significance of different MYC SVs in multiple myeloma has not been clearly established. EXPERIMENTAL DESIGN: We conducted a retrospective study of multiple myeloma comparing MYC SV subtypes identified by next-generation sequencing (NGS) and FISH to MYC expression and disease survival using 140 cases from Mayo Clinic and 658 cases from the MMRF CoMMpass study. RESULTS: MYC SVs were found in 41% of cases and were classified into nine subtypes. A correlation between the presence of a MYC SV and increased MYC expression was identified. Among the nine MYC subtypes, the non-immunoglobulin (non-Ig) insertion subtype was independently associated with improved outcomes, while the Ig insertion subtype, specifically involving the IgL gene partner, was independently associated with poorer outcomes compared with other MYC SV subtypes. Although the FISH methodology failed to detect approximately 70% of all MYC SVs, those detected by FISH were associated with elevated MYC gene expression and poor outcomes suggesting a different pathogenic role for FISH-detected MYC subtypes compared with other MYC subtypes. CONCLUSIONS: Understanding the impact of different MYC SVs on disease outcome is necessary for the reliable interpretation of MYC SVs in multiple myeloma. NGS approaches should be considered as a replacement technique for a more comprehensive evaluation of the multiple myeloma clone.

Inflation of tumor mutation burden by tumor-only sequencing in under-represented groups.

With the recent FDA approval of tumor mutational burden-high (TMB-H) status as a biomarker for treatment with a PD-1 inhibitor regardless of tumor type, accurate assessment of patient-specific TMB is more critical now more than ever. Using paired tumor and germline exome sequencing data from 701 patients newly diagnosed with multiple myeloma, including 575 self-reported White patients and 126 self-reported Black patients, we observed that compared to the gold standard of filtering germline variants with patient-paired germline sequencing data, TMB estimates were significantly higher in both Black and White patients when using public databases for filtering non-somatic mutations; however, TMB was more significantly inflated in Black patients compared to White patients. TMB as a biomarker for patient selection to receive immune checkpoint inhibitors (ICIs) therapy without patient-paired germline sequencing may introduce racial bias due to the under-representation of minority groups in public databases.

Human Cerebrospinal Fluid Modulates Pathways Promoting Glioblastoma Malignancy.

Glioblastoma (GBM) is the most common and devastating primary cancer of the central nervous system in adults. High grade gliomas are able to modify and respond to the brain microenvironment. When GBM tumors infiltrate the Subventricular zone (SVZ) they have a more aggressive clinical presentation than SVZ-distal tumors. We suggest that cerebrospinal fluid (CSF) contact contributes to enhance GBM malignant characteristics in these tumors. We evaluated the impact of human CSF on GBM, performing a transcriptome analysis on human primary GBM cells exposed to CSF to measure changes in gene expression profile and their clinical relevance on disease outcome. In addition we evaluated the proliferation and migration changes of CSF-exposed GBM cells in vitro and in vivo. CSF induced transcriptomic changes in pathways promoting cell malignancy, such as apoptosis, survival, cell motility, angiogenesis, inflammation, and glucose metabolism. A genetic signature extracted from the identified transcriptional changes in response to CSF proved to be predictive of GBM patient survival using the TCGA database. Furthermore, CSF induced an increase in viability, proliferation rate, and self-renewing capacity, as well as the migratory capabilities of GBM cells in vitro. In vivo, GBM cells co-injected with human CSF generated larger and more proliferative tumors compared to controls. Taken together, these results provide direct evidence that CSF is a key player in determining tumor growth and invasion through the activation of complex gene expression patterns characteristic of a malignant phenotype. These findings have diagnostic and therapeutic implications for GBM patients. The changes induced by CSF contact might play a role in the increased malignancy of SVZ-proximal GBM.

Alpha 1-antichymotrypsin contributes to stem cell characteristics and enhances tumorigenicity of glioblastoma.

BACKGROUND: Glioblastomas (GBMs) are the main primary brain tumors in adults with almost 100% recurrence rate. Patients with lateral ventricle proximal GBMs (LV-GBMs) exhibit worse survival compared to distal locations for unknown reasons. One hypothesis is the proximity of these tumors to the cerebrospinal fluid (CSF) and its chemical cues that can regulate cellular phenotype. We therefore investigated the role of CSF on GBM gene expression and the role of a CSF-induced gene, SERPINA3, in GBM malignancy in vitro and in vivo. METHODS: We utilized human CSF and GBM brain tumor-initiating cells (BTICs). We determined the impact of SERPINA3 expression in glioma patients using The Cancer Genome Atlas (TCGA) database. SERPINA3 expression changes were evaluated at mRNA and protein levels. The effects of knockdown (KD) and overexpression (OE) of SERPINA3 on cell migration, viability and cell proliferation were evaluated. Stem cell characteristics on KD cells were evaluated by differentiation and colony formation experiments. Tumor growth was studied by intracranial and flank injections. RESULTS: GBM-CSF increased BTIC migration accompanied by upregulation of the SERPINA3 gene. In patient samples and TCGA data, we observed SERPINA3 to correlate directly with brain tumor grade and indirectly with GBM patient survival. SERPINA3 KD induced a decrease in cell proliferation, migration, invasion, and stem cell characteristics, while SERPINA3 OE increased cell migration. In vivo, SERPINA3 KD BTICs showed increased survival in a murine model. CONCLUSIONS: SERPINA3 plays a key role in GBM malignancy and its inhibition results in a better outcome using GBM preclinical models.

Targeting of inflammatory pathways with R2CHOP in high-risk DLBCL.

Diffuse large B-cell lymphoma (DLBCL) is the most common lymphoma, and front line therapies have not improved overall outcomes since the advent of immunochemotherapy. By pairing DNA and gene expression data with clinical response data, we identified a high-risk subset of non-GCB DLBCL patients characterized by genomic alterations and expression signatures capable of sustaining an inflammatory environment. These mutational alterations (PIM1, SPEN, and MYD88 [L265P]) and expression signatures (NF-κB, IRF4, and JAK-STAT engagement) were associated with proliferative signaling, and were found to be enriched in patients treated with RCHOP that experienced unfavorable outcomes. However, patients with these high-risk mutations had more favorable outcomes when the immunomodulatory agent lenalidomide was added to RCHOP (R2CHOP). We are the first to report the genomic validation of a high-risk phenotype with a preferential response towards R2CHOP therapy in non-GCB DLBCL patients. These conclusions could be translated to a clinical setting to identify the ~38% of non-GCB patients that could be considered high-risk, and would benefit from alternative therapies to standard RCHOP based on personalized genomic data.

Monosomic loss of MIR15A/MIR16-1 is a driver of multiple myeloma proliferation and disease progression.

The most common genetic abnormality in multiple myeloma (MM) is the deletion of chromosome 13, seen in almost half of newly diagnosed patients. Unlike chronic lymphocytic leukemia, where a recurrent minimally deleted region including MIR15A/MIR16-1 has been mapped, the deletions in MM predominantly involve the entire chromosome and no specific driver gene has been identified. Additional candidate loci include RB1 and DIS3, but while biallelic deletion of RB1 is associated with disease progression, DIS3 is a common essential gene and complete inactivation is not observed. The Vk*MYC transgenic mouse model of MM spontaneously acquires del(14), syntenic to human chromosome 13, and Rb1 complete inactivation, but not Dis3 mutations. Taking advantage of this model, we explored the role in MM initiation and progression of two candidate loci on chromosome 13: RB1 and MIR15A/MIR16-1. Monoallelic deletion of Mir15a/Mir16-1 but not Rb1 was sufficient to accelerate the development of monoclonal gammopathy in wildtype mice, and the progression of MM in Vk*MYC mice, resulting in increased expression of Mir15a/Mir16-1 target genes and plasma cell proliferation, which was similarly observed in patients with MM.

Immune Responses and Risk of Triple-negative Breast Cancer: Implications for Higher Rates among African American Women.

The etiology of triple-negative breast cancers (TNBC) is poorly understood. As many TNBCs develop prior to the initiation of breast cancer screening or at younger ages when the sensitivity of mammography is comparatively low, understanding the etiology of TNBCs is critical for discovering novel prevention approaches for these tumors. Furthermore, the higher incidence rate of estrogen receptor-negative breast cancers, and specifically, of TNBCs, among young African American women (AAW) versus white women is a source of racial disparities in breast cancer mortality. Whereas immune responses to TNBCs have received considerable attention in relation to prognosis and treatment, the concept that dysregulated immune responses may predispose to the development of TNBCs has received limited attention. We present evidence that dysregulated immune responses are critical in the pathogenesis of TNBCs, based on the molecular biology of the cancers and the mechanisms proposed to mediate TNBC risk factors. Furthermore, proposed risk factors for TNBC, especially childbearing without breastfeeding, high parity, and obesity, are more prevalent among AAW than white women. Limited data suggest genetic differences in immune responses by race, which favor a stronger Thr type 2 (Th2) immune response among AAW than white women. Th2 responses contribute to wound-healing processes, which are implicated in the pathogenesis of TNBCs. Accordingly, we review data on the link between immune responses and TNBC risk and consider whether the prevalence of risk factors that result in dysregulated immunity is higher among AAW than white women.

HLA class-I and class-II restricted neoantigen loads predict overall survival in breast cancer.

Tumors acquire numerous mutations during development and progression. When translated into proteins, these mutations give rise to neoantigens that can be recognized by T cells and generate antibodies, representing an exciting direction of cancer immunotherapy. While neoantigens have been reported in many cancer types, the profiling of neoantigens often focused on the class-I subtype that are presented to CD8 + T cells, and the relationship between neoantigen load and clinical outcomes was often inconsistent among cancer types. In this study, we described an informatics workflow, REAL-neo, for identification, quality control (QC), and prioritization of both class-I and class-II human leukocyte antigen (HLA) bound neoantigens that arise from somatic single nucleotide mutations (SNM), small insertions and deletions (INDEL), and gene fusions. We applied REAL-neo to 835 primary breast tumors in the Cancer Genome Atlas (TCGA) and performed comprehensive profiling and characterization of the detected neoantigens. We found recurrent HLA class-I and class-II restricted neoantigens across breast cancer cases, and uncovered associations between neoantigen load and clinical traits. Both class-I and class-II neoantigen loads from SNM and INDEL were found to predict overall survival independent of tumor mutational burden (TMB), breast cancer subtypes, tumor-infiltrating lymphocyte (TIL) levels, tumor stage, and age at diagnosis. Our study highlighted the importance of accurate and comprehensive neoantigen profiling and QC, and is the first to report the predictive value of neoantigen load for overall survival in breast cancer.

Somatic selection distinguishes oncogenes and tumor suppressor genes.

MOTIVATION: Functions of cancer driver genes vary substantially across tissues and organs. Distinguishing passenger genes, oncogenes (OGs) and tumor-suppressor genes (TSGs) for each cancer type is critical for understanding tumor biology and identifying clinically actionable targets. Although many computational tools are available to predict putative cancer driver genes, resources for context-aware classifications of OGs and TSGs are limited. RESULTS: We show that the direction and magnitude of somatic selection of protein-coding mutations are significantly different for passenger genes, OGs and TSGs. Based on these patterns, we develop a new method (genes under selection in tumors) to discover OGs and TSGs in a cancer-type specific manner. Genes under selection in tumors shows a high accuracy (92%) when evaluated via strict cross-validations. Its application to 10 172 tumor exomes found known and novel cancer drivers with high tissue-specificities. In 11 out of 13 OGs shared among multiple cancer types, we found functional domains selectively engaged in different cancers, suggesting differences in disease mechanisms. AVAILABILITY AND IMPLEMENTATION: An R implementation of the GUST algorithm is available at https://github.com/liliulab/gust. A database with pre-computed results is available at https://liliulab.shinyapps.io/gust. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.