HLA-A*03 and response to immune checkpoint blockade in cancer: an epidemiological biomarker study.

BACKGROUND: Predictive biomarkers could allow more precise use of immune checkpoint inhibitors (ICIs) in treating advanced cancers. Given the central role of HLA molecules in immunity, variation at the HLA loci could differentially affect the response to ICIs. The aim of this epidemiological study was to determine the effect of HLA-A*03 as a biomarker for predicting response to immunotherapy. METHODS: In this epidemiological study, we investigated the clinical outcomes (overall survival, progression free survival, and objective response rate) after treatment for advanced cancer in eight cohorts of patients: three observational cohorts of patients with various types of advanced tumours (the Memorial Sloan Kettering Integrated Mutation Profiling of Actionable Cancer Targets [MSK-IMPACT] cohort, the Dana-Farber Cancer Institute [DFCI] Profile cohort, and The Cancer Genome Atlas) and five clinical trials of patients with advanced bladder cancer (JAVELIN Solid Tumour) or renal cell carcinoma (CheckMate-009, CheckMate-010, CheckMate-025, and JAVELIN Renal 101). In total, these cohorts included 3335 patients treated with various ICI agents (anti-PD-1, anti-PD-L1, and anti-CTLA-4 inhibitors) and 10 917 patients treated with non-ICI cancer-directed therapeutic approaches. We initially modelled the association of HLA amino-acid variation with overall survival in the MSK-IMPACT discovery cohort, followed by a detailed analysis of the association between HLA-A*03 and clinical outcomes in MSK-IMPACT, with replication in the additional cohorts (two further observational cohorts and five clinical trials). FINDINGS: HLA-A*03 was associated in an additive manner with reduced overall survival after ICI treatment in the MSK-IMPACT cohort (HR 1·48 per HLA-A*03 allele [95% CI 1·20-1·82], p=0·00022), the validation DFCI Profile cohort (HR 1·22 per HLA-A*03 allele, 1·05-1·42; p=0·0097), and in the JAVELIN Solid Tumour clinical trial for bladder cancer (HR 1·36 per HLA-A*03 allele, 1·01-1·85; p=0·047). The HLA-A*03 effect was observed across ICI agents and tumour types, but not in patients treated with alternative therapies. Patients with HLA-A*03 had shorter progression-free survival in the pooled patient population from the three CheckMate clinical trials of nivolumab for renal cell carcinoma (HR 1·31, 1·01-1·71; p=0·044), but not in those receiving control (everolimus) therapies. Objective responses were observed in none of eight HLA-A*03 homozygotes in the ICI group (compared with 59 [26·6%] of 222 HLA-A*03 non-carriers and 13 (17·1%) of 76 HLA-A*03 heterozygotes). HLA-A*03 was associated with shorter progression-free survival in patients receiving ICI in the JAVELIN Renal 101 randomised clinical trial for renal cell carcinoma (avelumab plus axitinib; HR 1·59 per HLA-A*03 allele, 1·16-2·16; p=0·0036), but not in those receiving control (sunitinib) therapy. Objective responses were recorded in one (12·5%) of eight HLA-A*03 homozygotes in the ICI group (compared with 162 [63·8%] of 254 HLA-A*03 non-carriers and 40 [55·6%] of 72 HLA-A*03 heterozygotes). HLA-A*03 was associated with impaired outcome in meta-analysis of all 3335 patients treated with ICI at genome-wide significance (p=2·01 × 10-8) with no evidence of heterogeneity in effect (I2 0%, 95% CI 0-0·76) INTERPRETATION: HLA-A*03 is a predictive biomarker of poor response to ICI. Further evaluation of HLA-A*03 is warranted in randomised trials. HLA-A*03 carriage could be considered in decisions to initiate ICI in patients with cancer. FUNDING: National Institutes of Health, Merck KGaA, and Pfizer.

Cross-talk between chromatin acetylation and SUMOylation of tripartite motif-containing protein 24 (TRIM24) impacts cell adhesion.

Proteins with domains that recognize and bind post-translational modifications (PTMs) of histones are collectively termed epigenetic readers. Numerous interactions between specific reader protein domains and histone PTMs and their regulatory outcomes have been reported, but little is known about how reader proteins may in turn be modulated by these interactions. Tripartite motif-containing protein 24 (TRIM24) is a histone reader aberrantly expressed in multiple cancers. Here, our investigation revealed functional cross-talk between histone acetylation and TRIM24 SUMOylation. Binding of TRIM24 to chromatin via its tandem PHD-bromodomain, which recognizes unmethylated lysine 4 and acetylated lysine 23 of histone H3 (H3K4me0/K23ac), led to TRIM24 SUMOylation at lysine residues 723 and 741. Inactivation of the bromodomain, either by mutation or with a small-molecule inhibitor, IACS-9571, abolished TRIM24 SUMOylation. Conversely, inhibition of histone deacetylation markedly increased TRIM24's interaction with chromatin and its SUMOylation. Of note, gene expression profiling of MCF7 cells expressing WT versus SUMO-deficient TRIM24 identified cell adhesion as the major pathway regulated by the cross-talk between chromatin acetylation and TRIM24 SUMOylation. In conclusion, our findings establish a new link between histone H3 acetylation and SUMOylation of the reader protein TRIM24, a functional connection that may bear on TRIM24's oncogenic function and may inform future studies of PTM cross-talk between histones and epigenetic regulators.

A cis-regulatory map of the Drosophila genome.

Systematic annotation of gene regulatory elements is a major challenge in genome science. Direct mapping of chromatin modification marks and transcriptional factor binding sites genome-wide has successfully identified specific subtypes of regulatory elements. In Drosophila several pioneering studies have provided genome-wide identification of Polycomb response elements, chromatin states, transcription factor binding sites, RNA polymerase II regulation and insulator elements; however, comprehensive annotation of the regulatory genome remains a significant challenge. Here we describe results from the modENCODE cis-regulatory annotation project. We produced a map of the Drosophila melanogaster regulatory genome on the basis of more than 300 chromatin immunoprecipitation data sets for eight chromatin features, five histone deacetylases and thirty-eight site-specific transcription factors at different stages of development. Using these data we inferred more than 20,000 candidate regulatory elements and validated a subset of predictions for promoters, enhancers and insulators in vivo. We identified also nearly 2,000 genomic regions of dense transcription factor binding associated with chromatin activity and accessibility. We discovered hundreds of new transcription factor co-binding relationships and defined a transcription factor network with over 800 potential regulatory relationships.

Computational characterisation of cancer molecular profiles derived using next generation sequencing.

Our current understanding of cancer genetics is grounded on the principle that cancer arises from a clone that has accumulated the requisite somatically acquired genetic aberrations, leading to the malignant transformation. It also results in aberrent of gene and protein expression. Next generation sequencing (NGS) or deep sequencing platforms are being used to create large catalogues of changes in copy numbers, mutations, structural variations, gene fusions, gene expression, and other types of information for cancer patients. However, inferring different types of biological changes from raw reads generated using the sequencing experiments is algorithmically and computationally challenging. In this article, we outline common steps for the quality control and processing of NGS data. We highlight the importance of accurate and application-specific alignment of these reads and the methodological steps and challenges in obtaining different types of information. We comment on the importance of integrating these data and building infrastructure to analyse it. We also provide exhaustive lists of available software to obtain information and point the readers to articles comparing software for deeper insight in specialised areas. We hope that the article will guide readers in choosing the right tools for analysing oncogenomic datasets.

Lessons from a decade of integrating cancer copy number alterations with gene expression profiles.

Over the last decade, multiple functional genomic datasets studying chromosomal aberrations and their downstream effects on gene expression have accumulated for several cancer types. A vast majority of them are in the form of paired gene expression profiles and somatic copy number alterations (CNA) information on the same patients identified using microarray platforms. In response, many algorithms and software packages are available for integrating these paired data. Surprisingly, there has been no serious attempt to review the currently available methodologies or the novel insights brought using them. In this work, we discuss the quantitative relationships observed between CNA and gene expression in multiple cancer types and biological milestones achieved using the available methodologies. We discuss the conceptual evolution of both, the step-wise and the joint data integration methodologies over the last decade. We conclude by providing suggestions for building efficient data integration methodologies and asking further biological questions.

Integrative analysis of gene and miRNA expression profiles with transcription factor-miRNA feed-forward loops identifies regulators in human cancers.

We describe here a novel method for integrating gene and miRNA expression profiles in cancer using feed-forward loops (FFLs) consisting of transcription factors (TFs), miRNAs and their common target genes. The dChip-GemiNI (Gene and miRNA Network-based Integration) method statistically ranks computationally predicted FFLs by their explanatory power to account for differential gene and miRNA expression between two biological conditions such as normal and cancer. GemiNI integrates not only gene and miRNA expression data but also computationally derived information about TF-target gene and miRNA-mRNA interactions. Literature validation shows that the integrated modeling of expression data and FFLs better identifies cancer-related TFs and miRNAs compared to existing approaches. We have utilized GemiNI for analyzing six data sets of solid cancers (liver, kidney, prostate, lung and germ cell) and found that top-ranked FFLs account for ∼20% of transcriptome changes between normal and cancer. We have identified common FFL regulators across multiple cancer types, such as known FFLs consisting of MYC and miR-15/miR-17 families, and novel FFLs consisting of ARNT, CREB1 and their miRNA partners. The results and analysis web server are available at http://www.canevolve.org/dChip-GemiNi.

Avelumab first-line maintenance in advanced urothelial carcinoma: Complete screening for prognostic and predictive factors using machine learning in the JAVELIN Bladder 100 phase 3 trial.

BACKGROUND: Avelumab first-line (1 L) maintenance is a standard of care for advanced urothelial carcinoma (aUC) based on the JAVELIN Bladder 100 phase 3 trial, which showed that avelumab 1 L maintenance + best supportive care (BSC) significantly prolonged overall survival (OS) and progression-free survival (PFS) vs BSC alone in patients who were progression free after receiving 1 L platinum-containing chemotherapy. Here, we comprehensively screened JAVELIN Bladder 100 trial datasets to identify prognostic factors that define subpopulations of patients with longer or shorter OS irrespective of treatment, and predictive factors that select patients who could obtain a greater OS benefit from avelumab 1 L maintenance treatment. METHODS: We performed machine learning analyses to screen a large set of baseline covariates, including patient demographics, disease characteristics, laboratory values, molecular biomarkers, and patient-reported outcomes. Covariates were identified from previously reported analyses and established prognostic and predictive markers. Variables selected from random survival forest models were processed further in univariate Cox models with treatment interaction and visually inspected using correlation analysis and Kaplan-Meier curves. Results were summarized in a multivariable Cox model. RESULTS: Prognostic baseline covariates associated with OS included in the final model were assignment to avelumab 1 L maintenance treatment, Eastern Cooperative Oncology Group performance status, site of metastasis, sum of longest target lesion diameters, levels of C-reactive protein and alkaline phosphatase in blood, lymphocyte proportion in intratumoral stroma, tumor mutational burden, and tumor CD8+ T-cell infiltration. Potential predictive factors included site of metastasis, tumor mutation burden, and tumor CD8+ T-cell infiltration. An analysis in patients with PD-L1+ tumors had similar findings to those in the overall population. CONCLUSIONS: Machine learning analyses of data from the JAVELIN Bladder 100 trial identified potential prognostic and predictive factors for avelumab 1 L maintenance treatment in patients with aUC, which warrant further evaluation in other clinical datasets.

Biomarker Analyses Investigating Disease Biology and Associations with Outcomes in the JAVELIN Merkel 200 Trial of Avelumab in Metastatic Merkel Cell Carcinoma.

PURPOSE: Avelumab (anti-PD-L1) became the first approved treatment for metastatic Merkel cell carcinoma (mMCC) based on results from the phase II JAVELIN Merkel 200 trial. In this study, we report exploratory biomarker analyses from the trial. PATIENTS AND METHODS: Patients with mMCC (n = 88) with or without prior first-line chemotherapy received avelumab 10 mg/kg every 2 weeks. We conducted analyses on somatic mutations, mutational signatures, and tumor mutational burden using paired whole-exome sequencing. Additionally, we examined gene and gene set expression, immune content from RNA sequencing profiles, as well as tumor PD-L1 and CD8 statuses from IHC and CD8 status from digital pathology. RESULTS: Tumors positive for Merkel cell polyomavirus (MCPyV) were characterized by an absence of driver mutations and a low tumor mutational burden, consistent with previous studies. A novel MCPyV-specific host gene expression signature was identified. MCPyV+ tumors had increased levels of immunosuppressive M2 macrophages in the tumor microenvironment, which seemed to correlate with PD-L1 expression; high CD8+ T-cell density in these tumors did not predict response to avelumab. Conversely, in patients with MCPyV- tumors, higher CD8+ T-cell density seemed to be associated with response to avelumab. Mutations in several genes were associated with treatment outcomes. Compared with tumors sampled before chemotherapy, tumors sampled after chemotherapy had downregulated gene signatures for immune responses, including reduced expression of IFNγ-related pathways. Levels of activated dendritic cells in responding patients were higher in patients assessed after versus before chemotherapy. CONCLUSIONS: Exploratory analyses provide insights into mMCC biology and potential associations with response to avelumab. Chemotherapy seems to negatively modulate the immune microenvironment.

The shaping and functional consequences of the dosage effect landscape in multiple myeloma.

BACKGROUND: Multiple myeloma (MM) is a malignant proliferation of plasma B cells. Based on recurrent aneuploidy such as copy number alterations (CNAs), myeloma is divided into two subtypes with different CNA patterns and patient survival outcomes. How aneuploidy events arise, and whether they contribute to cancer cell evolution are actively studied. The large amount of transcriptomic changes resultant of CNAs (dosage effect) pose big challenges for identifying functional consequences of CNAs in myeloma in terms of specific driver genes and pathways. In this study, we hypothesize that gene-wise dosage effect varies as a result from complex regulatory networks that translate the impact of CNAs to gene expression, and studying this variation can provide insights into functional effects of CNAs. RESULTS: We propose gene-wise dosage effect score and genome-wide karyotype plot as tools to measure and visualize concordant copy number and expression changes across cancer samples. We find that dosage effect in myeloma is widespread yet variable, and it is correlated with gene expression level and CNA frequencies in different chromosomes. Our analysis suggests that despite the enrichment of differentially expressed genes between hyperdiploid MM and non-hyperdiploid MM in the trisomy chromosomes, the chromosomal proportion of dosage sensitive genes is higher in the non-trisomy chromosomes. Dosage-sensitive genes are enriched by genes with protein translation and localization functions, and dosage resistant genes are enriched by apoptosis genes. These results point to future studies on differential dosage sensitivity and resistance of pro- and anti-proliferation pathways and their variation across patients as therapeutic targets and prognosis markers. CONCLUSIONS: Our findings support the hypothesis that recurrent CNAs in myeloma are selected by their functional consequences. The novel dosage effect score defined in this work will facilitate integration of copy number and expression data for identifying driver genes in cancer genomics studies. The accompanying R code is available at http://www.canevolve.org/dosageEffect/.

A comprehensive map of insulator elements for the Drosophila genome.

Insulators are DNA sequences that control the interactions among genomic regulatory elements and act as chromatin boundaries. A thorough understanding of their location and function is necessary to address the complexities of metazoan gene regulation. We studied by ChIP-chip the genome-wide binding sites of 6 insulator-associated proteins-dCTCF, CP190, BEAF-32, Su(Hw), Mod(mdg4), and GAF-to obtain the first comprehensive map of insulator elements in Drosophila embryos. We identify over 14,000 putative insulators, including all classically defined insulators. We find two major classes of insulators defined by dCTCF/CP190/BEAF-32 and Su(Hw), respectively. Distributional analyses of insulators revealed that particular sub-classes of insulator elements are excluded between cis-regulatory elements and their target promoters; divide differentially expressed, alternative, and divergent promoters; act as chromatin boundaries; are associated with chromosomal breakpoints among species; and are embedded within active chromatin domains. Together, these results provide a map demarcating the boundaries of gene regulatory units and a framework for understanding insulator function during the development and evolution of Drosophila.

Determinants of antigenicity and specificity in immune response for protein sequences.

BACKGROUND: Target specific antibodies are pivotal for the design of vaccines, immunodiagnostic tests, studies on proteomics for cancer biomarker discovery, identification of protein-DNA and other interactions, and small and large biochemical assays. Therefore, it is important to understand the properties of protein sequences that are important for antigenicity and to identify small peptide epitopes and large regions in the linear sequence of the proteins whose utilization result in specific antibodies. RESULTS: Our analysis using protein properties suggested that sequence composition combined with evolutionary information and predicted secondary structure, as well as solvent accessibility is sufficient to predict successful peptide epitopes. The antigenicity and the specificity in immune response were also found to depend on the epitope length. We trained the B-Cell Epitope Oracle (BEOracle), a support vector machine (SVM) classifier, for the identification of continuous B-Cell epitopes with these protein properties as learning features. The BEOracle achieved an F1-measure of 81.37% on a large validation set. The BEOracle classifier outperformed the classical methods based on propensity and sophisticated methods like BCPred and Bepipred for B-Cell epitope prediction. The BEOracle classifier also identified peptides for the ChIP-grade antibodies from the modENCODE/ENCODE projects with 96.88% accuracy. High BEOracle score for peptides showed some correlation with the antibody intensity on Immunofluorescence studies done on fly embryos. Finally, a second SVM classifier, the B-Cell Region Oracle (BROracle) was trained with the BEOracle scores as features to predict the performance of antibodies generated with large protein regions with high accuracy. The BROracle classifier achieved accuracies of 75.26-63.88% on a validation set with immunofluorescence, immunohistochemistry, protein arrays and western blot results from Protein Atlas database. CONCLUSIONS: Together our results suggest that antigenicity is a local property of the protein sequences and that protein sequence properties of composition, secondary structure, solvent accessibility and evolutionary conservation are the determinants of antigenicity and specificity in immune response. Moreover, specificity in immune response could also be accurately predicted for large protein regions without the knowledge of the protein tertiary structure or the presence of discontinuous epitopes. The dataset prepared in this work and the classifier models are available for download at https://sites.google.com/site/oracleclassifiers/.

The dChip survival analysis module for microarray data.

BACKGROUND: Genome-wide expression signatures are emerging as potential marker for overall survival and disease recurrence risk as evidenced by recent commercialization of gene expression based biomarkers in breast cancer. Similar predictions have recently been carried out using genome-wide copy number alterations and microRNAs. Existing software packages for microarray data analysis provide functions to define expression-based survival gene signatures. However, there is no software that can perform survival analysis using SNP array data or draw survival curves interactively for expression-based sample clusters. RESULTS: We have developed the survival analysis module in the dChip software that performs survival analysis across the genome for gene expression and copy number microarray data. Built on the current dChip software's microarray analysis functions such as chromosome display and clustering, the new survival functions include interactive exploring of Kaplan-Meier (K-M) plots using expression or copy number data, computing survival p-values from the log-rank test and Cox models, and using permutation to identify significant chromosome regions associated with survival. CONCLUSIONS: The dChip survival module provides user-friendly way to perform survival analysis and visualize the results in the context of genes and cytobands. It requires no coding expertise and only minimal learning curve for thousands of existing dChip users. The implementation in Visual C++ also enables fast computation. The software and demonstration data are freely available at http://dchip-surv.chenglilab.org.

First-line avelumab in a cohort of 116 patients with metastatic Merkel cell carcinoma (JAVELIN Merkel 200): primary and biomarker analyses of a phase II study.

BACKGROUND: Avelumab (anti-programmed death ligand 1 (PD-L1)) is approved in multiple countries for the treatment of metastatic Merkel cell carcinoma (mMCC), a rare and aggressive skin cancer. We report efficacy and safety data and exploratory biomarker analyses from a cohort of patients with mMCC treated with first-line avelumab in a phase II trial. METHODS: Patients with treatment-naive mMCC received avelumab 10 mg/kg intravenously every 2 weeks. The primary endpoint was durable response, defined as objective response (complete or partial response; assessed by independent review) lasting ≥6 months. Additional assessments included progression-free survival (PFS), overall survival (OS), safety, and biomarker analyses. RESULTS: In 116 patients treated with avelumab, median follow-up was 21.2 months (range: 14.9-36.6). Thirty-five patients had a response lasting ≥6 months, giving a durable response rate of 30.2% (95% CI: 22.0% to 39.4%). The objective response rate was 39.7% (95% CI: 30.7% to 49.2%). Median PFS was 4.1 months (95% CI: 1.4 to 6.1) and median OS was 20.3 months (95% CI: 12.4 to not estimable). Response rates were numerically higher in patients with PD-L1+ tumors, Merkel cell polyomavirus (MCPyV)-negative tumors, and tumors with increased intratumoral CD8+ T-cell density. Exploratory analyses did not identify a biomarker that could reliably predict a response to first-line treatment with avelumab; however, a novel gene expression signature to identify the presence of MCPyV+ tumors was derived. Treatment-related adverse events (any grade) occurred in 94 (81.0%) patients, including grade 3/4 events in 21 (18.1%) patients; no treatment-related deaths occurred. CONCLUSION: In patients with mMCC, first-line treatment with avelumab led to responses in 40% and durable responses in 30%, and was associated with a low rate of grade 3/4 treatment-related adverse events.

Flynet: a genomic resource for Drosophila melanogaster transcriptional regulatory networks.

MOTIVATION: The highly coordinated expression of thousands of genes in an organism is regulated by the concerted action of transcription factors, chromatin proteins and epigenetic mechanisms. High-throughput experimental data for genome wide in vivo protein-DNA interactions and epigenetic marks are becoming available from large projects, such as the model organism ENCyclopedia Of DNA Elements (modENCODE) and from individual labs. Dissemination and visualization of these datasets in an explorable form is an important challenge. RESULTS: To support research on Drosophila melanogaster transcription regulation and make the genome wide in vivo protein-DNA interactions data available to the scientific community as a whole, we have developed a system called Flynet. Currently, Flynet contains 101 datasets for 38 transcription factors and chromatin regulator proteins in different experimental conditions. These factors exhibit different types of binding profiles ranging from sharp localized peaks to broad binding regions. The protein-DNA interaction data in Flynet was obtained from the analysis of chromatin immunoprecipitation experiments on one color and two color genomic tiling arrays as well as chromatin immunoprecipitation followed by massively parallel sequencing. A web-based interface, integrated with an AJAX based genome browser, has been built for queries and presenting analysis results. Flynet also makes available the cis-regulatory modules reported in literature, known and de novo identified sequence motifs across the genome, and other resources to study gene regulation. AVAILABILITY: Flynet is available at https://www.cistrack.org/flynet/.

Avelumab in patients with previously treated metastatic Merkel cell carcinoma: long-term data and biomarker analyses from the single-arm phase 2 JAVELIN Merkel 200 trial.

BACKGROUND: Merkel cell carcinoma (MCC) is a rare, aggressive skin cancer associated with a high risk of metastasis. In 2017, avelumab (anti-programmed death-ligand 1 (PD-L1)) became the first approved treatment for patients with metastatic MCC (mMCC), based on the occurrence of durable responses in a subset of patients. Here, we report long-term efficacy and safety data and exploratory biomarker analyses in patients with mMCC treated with avelumab. METHODS: In a cohort of this single-arm, phase 2 trial (JAVELIN Merkel 200), patients with mMCC and disease progression after prior chemotherapy received avelumab 10 mg/kg intravenously every 2 weeks. The primary endpoint was confirmed objective response rate (ORR) by independent review per Response Evaluation Criteria in Solid Tumors V.1.1. Other assessments included duration of response, progression-free survival, overall survival (OS), safety and biomarker analyses. RESULTS: As of 14 September 2018, 88 patients had been followed up for a median of 40.8 months (range 36.4-49.7 months). The ORR was 33.0% (95% CI 23.3% to 43.8%), including a complete response in 11.4% (10 patients), and the median duration of response was 40.5 months (95% CI 18.0 months to not estimable). As of 2 May 2019 (≥44 months of follow-up), the median OS was 12.6 months (95% CI 7.5 to 17.1 months) and the 42-month OS rate was 31% (95% CI 22% to 41%). Of long-term survivors (OS >36 months) evaluable for PD-L1 expression status (n=22), 81.8% had PD-L1+ tumors. In exploratory biomarker analyses, high tumor mutational burden (≥2 non-synonymous somatic variants per megabase) and high major histocompatibility complex class I expression (30% of tumors with highest expression) were associated with trends for improved ORR and OS. In long-term safety assessments (≥36 months of follow-up), no new or unexpected adverse events were reported, and no treatment-related deaths occurred. CONCLUSIONS: Avelumab showed continued durable responses and meaningful long-term survival outcomes in patients with mMCC, reinforcing avelumab as a standard-of-care treatment option for this disease. TRIAL REGISTRATION NUMBER: NCT02155647.

Computational characterization of multiple Gag-like human proteins.

In a genome-wide analysis, we have identified 85 human genes encoding 103 protein isoforms that resemble retroviral Gag proteins. These genes were domesticated from retrotransposons in at least five independent events during vertebrate evolution and were subsequently duplicated further in mammals. Structural insights into the mammalian proteins can be inferred by homology to Gag from viruses such as HIV; in turn, the cellular roles of the mammalian Gag homologs, such as apoptosis-related functions and binding to ubiquitin ligases, might hint at further functionality of viral Gag itself.

Genomic analysis of estrogen cascade reveals histone variant H2A.Z associated with breast cancer progression.

We demonstrate an integrated approach to the study of a transcriptional regulatory cascade involved in the progression of breast cancer and we identify a protein associated with disease progression. Using chromatin immunoprecipitation and genome tiling arrays, whole genome mapping of transcription factor-binding sites was combined with gene expression profiling to identify genes involved in the proliferative response to estrogen (E2). Using RNA interference, selected ERalpha and c-MYC gene targets were knocked down to identify mediators of E2-stimulated cell proliferation. Tissue microarray screening revealed that high expression of an epigenetic factor, the E2-inducible histone variant H2A.Z, is significantly associated with lymph node metastasis and decreased breast cancer survival. Detection of H2A.Z levels independently increased the prognostic power of biomarkers currently in clinical use. This integrated approach has accelerated the identification of a molecule linked to breast cancer progression, has implications for diagnostic and therapeutic interventions, and can be applied to a wide range of cancers.

Applications of machine learning in drug discovery and development.

Drug discovery and development pipelines are long, complex and depend on numerous factors. Machine learning (ML) approaches provide a set of tools that can improve discovery and decision making for well-specified questions with abundant, high-quality data. Opportunities to apply ML occur in all stages of drug discovery. Examples include target validation, identification of prognostic biomarkers and analysis of digital pathology data in clinical trials. Applications have ranged in context and methodology, with some approaches yielding accurate predictions and insights. The challenges of applying ML lie primarily with the lack of interpretability and repeatability of ML-generated results, which may limit their application. In all areas, systematic and comprehensive high-dimensional data still need to be generated. With ongoing efforts to tackle these issues, as well as increasing awareness of the factors needed to validate ML approaches, the application of ML can promote data-driven decision making and has the potential to speed up the process and reduce failure rates in drug discovery and development.

Enhanced function annotations for Drosophila serine proteases: a case study for systematic annotation of multi-member gene families.

Systematically annotating function of enzymes that belong to large protein families encoded in a single eukaryotic genome is a very challenging task. We carried out such an exercise to annotate function for serine-protease family of the trypsin fold in Drosophila melanogaster, with an emphasis on annotating serine-protease homologues (SPHs) that may have lost their catalytic function. Our approach involves data mining and data integration to provide function annotations for 190 Drosophila gene products containing serine-protease-like domains, of which 35 are SPHs. This was accomplished by analysis of structure-function relationships, gene-expression profiles, large-scale protein-protein interaction data, literature mining and bioinformatic tools. We introduce functional residue clustering (FRC), a method that performs hierarchical clustering of sequences using properties of functionally important residues and utilizes correlation co-efficient as a quantitative similarity measure to transfer in vivo substrate specificities to proteases. We show that the efficiency of transfer of substrate-specificity information using this method is generally high. FRC was also applied on Drosophila proteases to assign putative competitive inhibitor relationships (CIRs). Microarray gene-expression data were utilized to uncover a large-scale and dual involvement of proteases in development and in immune response. We found specific recruitment of SPHs and proteases with CLIP domains in immune response, suggesting evolution of a new function for SPHs. We also suggest existence of separate downstream protease cascades for immune response against bacterial/fungal infections and parasite/parasitoid infections. We verify quality of our annotations using information from RNAi screens and other evidence types. Utilization of such multi-fold approaches results in 10-fold increase of function annotation for Drosophila serine proteases and demonstrates value in increasing annotations in multiple genomes.