Search

Celem niniejszego artykułu jest przeanalizowanie zależności między rentownością banków w Indiach a ich strukturą własności. Powody dokonania pomiaru siły oddziaływania struktury własności wyłoniły się z teorii zawodności rządu. Wykorzystano niezależnie skonstruowany zbiór danych obejmujący wszystkie komercyjne banki sektora publicznego i prywatnego w Indiach według stanu z kwietnia 2020 r. Dane dotyczą okresu od 2004 do 2020 r. Poszczególne dane o bankach zebrano z ich stron internetowych, a hipotezy testowano poprzez oszacowanie modelu ekonometrycznego, w tym przypadku modelu OLS typu pooled. Wnioskiem płynącym z badania jest to, że wyniki banków będących własnością rządu są gorsze w porównaniu z tymi uzyskiwanymi przez banki prywatne. Przyczynami mogą być: ogromna liczba kredytów sankcjonowanych w sektorach priorytetowych, nieuczciwe praktyki wynikające z działania grup interesów, korupcja oraz nieefektywność pracowników w sektorach publicznych.

Enrichment of Immune signaling in the integrated pan-cancer cluster. Different level of evidence support the association of the integrated pan-cancer cluster with an immune phenotype. (A) Enrichment of T- and B-cell signaling shown on the integrated map, yellow gradient (see Supplemental Methods). (B) Enrichment of high ESTIMATE scores in the pan-cancer samples, waterfall plot with pan-cancer cluster samples in red. (C) Enrichment of the immune-related pathways identified by differential expression analysis. (D) Unsupervised analysis of master regulator activities inferred by the MARINa method. Gene clusters enriched for T- and B-cell signaling, interferon signaling, and TNFA via NFKB signaling, red font. (E) The top enriched pathways based on the output of master regulator scores derived with MARINa are immune-related. (F) Copy number events in the integrated pan-cancer cluster compared to other samples in the full cohort. The pan-cancer cluster shows a lower number of copy number events in both arm-level events and focal events. Arm-level events (pan-cancer group on the right, background cohort on the left). (G) Same as F but focal events (pan-cancer group on the right, background cohort on the left). (H) Purity estimates in the integrated pan-cancer cluster compared to all other samples in the cohort. The pan-cancer cluster (green box) shows lower purity when compared to the whole cohort as a background (yellow box). This finding is consistent with other analyses indicating high immune signaling in the pan-cancer cluster.

Supplemental Figure 1: RSS transformed correlations correct for platform-specific inflation biases. Maps produced from different molecular data types. Samples in the maps are colored by tissue of origin. (A) Similarity score distributions for every pair of samples for each individual data platform. Spearman Rho was computed for mRNA, miRNA, RPPA, SCNV, and methylation platforms. HOCUS score was used for mutation data. (B) RSS-standardized similarity scores for every pair of samples for each individual data platform. The density curves demonstrate that RSS transformation corrects for platform-specific inflation biases. (C) (i)-(vi) Maps produced from each of the six molecular data types using OpenOrd layout. (vii) mRNA expression map produced using Principal Component Analysis method. Principal components 1 and 2 are used to represent the mRNA expression space. (viii) mRNA expression map produced using tSNE method. Dimensions 1 and 2 are used to represent the mRNA expression space. (D) Maps produced from inferred gene activities using the PARADIGM and SPIA methods. (E) Maps integrating more than one molecular data type.

Molecular subtype visualization. Known biology recapitulated by the mRNA expression map. (A) BRCA molecular subtypes and their layout in the map. The map shows that tumors of the same molecular subtype tend to cluster together, with very little mixing of those subtypes, indicating that those subtypes are indeed molecularly different. (B) Estrogen signaling (yellow) and basal signaling (blue) are the top differentiating programs between the basal and non-basal BRCA tumors. Very few tumors exhibit signals of both programs (green). The group of samples labeled as Basal subtype in part A predominantly exhibits the basal signaling program and the group of samples consisting of LumA (for luminal A) and LumB (for luminal B) in part A predominantly exhibits the estrogen signaling program. (C) Association of TP53 mutations (blue) with basal breast tumors and PIK3CA mutations (yellow) with luminal tumors demonstrates at-a-glance view of genomic event association with molecular subtypes in the map. Samples that exhibit both mutations are shown in green. (D) COAD and READ tumors cluster together (left) and the map separates genomically stable and unstable tumors (right). (E) BLCA tumors separate into three previously discovered molecular subtypes. These subtypes are BLCA-core, BLCA-lung-like, and BLCA-squamous-like. (F) KIRC tumors are deficient in MSH2 (activity level indicated by the intensity of yellow), a component of the DNA mismatch repair pathway. This is a known characterization of KIRC tumors. (G) Co-occurrence of VHL mutations (green) and high HIF1A activity (indicated by the intensity of yellow) in KIRC tumors. Samples colored in yellow indicate an absence of VHL mutation but high HIF1A activity. Samples colored in green indicate both the presence of VHL mutation and high HIF1A activity. No samples show a presence of VHL mutation and low HIF1A activity (such samples would be colored in blue). (H) Spatial correlation analysis (SCA). Mutual exclusivity of RB1 mutations and CDKN2A deletions. These two events do not co-occur in the same patients even when they occur in patients clustering in the same regions of the map. (I) SCA - positive spatial correlation of RB1 mutations and PTEN mutations. In regions of the map where both of these events occur, they tend to occur in the same patients. (J) The overlap in gene signatures distinctive of the Tumor Map's relative positioning of BRCA molecular subtypes. HER2+ tumors share more differentially expressed genes in common between luminal than basal tumors. The subtype groupings are defined by BRCA sample annotations. Differential expression analysis resulted in 150-gene signatures for each subtype. Venn-diagram of these gene signatures shows the overlap between gene sets for each of the subtype.

More detail into the methodology behind the resource are provided as well as proof-of-concept benchmarks.

Vast amounts of molecular data are being collected on tumor samples, which provide unique opportunities for discovering trends within and between cancer subtypes. Such cross-cancer analyses require computational methods that enable intuitive and interactive browsing of thousands of samples based on their molecular similarity. We created a portal called TumorMap to assist in exploration and statistical interrogation of high-dimensional complex “omics” data in an interactive and easily interpretable way. In the TumorMap, samples are arranged on a hexagonal grid based on their similarity to one another in the original genomic space and are rendered with Google's Map technology. While the important feature of this public portal is the ability for the users to build maps from their own data, we pre-built genomic maps from several previously published projects. We demonstrate the utility of this portal by presenting results obtained from The Cancer Genome Atlas project data. Cancer Res; 77(21); e111–4. ©2017 AACR.

Video showing basic functionality of the TumorMap for the first-time user.

Supplemental Table 1A: Differential expression scores used in the pan-cancer cluster GSEA analysis. These scores represent integrated t-statistic scores from the differential analysis within each tissue (see Supplemental Methods). Supplemental Table 1B: Full results of the GSEA analysis of the pan-cancer cluster, based on differential expression scores. Supplemental Table 1C: Per-sample ESTIMATE scores used in analysis of the pan-cancer cluster. Supplemental Table 1D: Per-sample surrogate purity scores used in analysis of the pan-cancer cluster. Supplemental Table 1E: List of genes located on sex chromosomes that were excluded from the methylation dataset in order to analyze only autosomal gene features. Supplemental Table 1F: Attributes available in the TumorMap that annotate metadata of the samples, along with descriptions of those attributes. Supplemental Table 1G: Statistical tests computed by different attribute enrichment analysis (AEA) tools available in the TumorMap. Supplemental Table 1H: List of 82 samples in the pan-cancer cluster in the integrated map as well as tissue composition, along with the number of samples, in the integrated pan-cancer cluster. Supplemental Table 1I: Input data for the LAML survival analysis.

Supplemental Figure 5: (A) Overview of the method to compute differential expression for samples in the integrated pan-cancer group vs. other samples in the TCGA cohort. Tissue composition imbalance was corrected for by performing t-tests within each tissue. For each gene, t-statistics were computed within each tumor type separately and then summarized per-gene t-statistics were calculated as an arithmetic mean, weighted by the inverse variance of the all the tissue-specific t-statistics values. (B) The distribution of log-transformed TPM values estimated from the mRNA-Seq data showing a normal-like distribution. (C) Comparing the log-transformed TPM values of the mRNA-Seq data to a normal distribution shows a reasonable agreement with the tumor data having slightly heavier tails. (D) Venn diagram representing the innate and adaptive immune systems and different levels of evidence supporting higher activity of each of the components of those systems in the pan-cancer cluster when compared to the rest of the TCGA cohort. We found evidence of both innate and adaptive immune system signaling with a number of different analyses. (E) Detailed version of the immune-related pathway characterizing the integrated pan-cancer cluster. This version of the pathway shows additional genes and their connections that are not shown in the summarized version of the immune signaling pathway of Figure 1D of the main text.

Tissue and molecular subtype distribution among the samples in the entire cohort (A-B), and in the pan-cancer cluster (C-D). The pie charts represent the number of samples from each tissue of origin in the entire cohort (A) and the integrated pan-cancer cluster (C). Black and white matrices illustrate the presence of molecular features of each platform (x-axis) across samples (y-axis), in the entire cohort (B) or in the integrated pan-cancer cluster (D). Data available for this sample for a given platform is marked black, otherwise the entry is white.

IRE1α-XBP1 signaling is emerging as a central orchestrator of malignant progression and immunosuppression in various cancer types. Employing a computational XBP1s detection method applied to TCGA datasets, we demonstrate that expression of the XBP1s mRNA isoform predicts poor survival in non-small cell lung cancer (NSCLC) patients. Ablation of IRE1α in malignant cells delays tumor progression and extends survival in mouse models of NSCLC. This protective effect is accompanied by alterations in intratumoral immune cell subsets eliciting durable adaptive anti-cancer immunity. Mechanistically, cancer cell-intrinsic IRE1α activation sustains mPGES-1 expression, enabling production of the immunosuppressive lipid mediator prostaglandin E2. Accordingly, restoring mPGES-1 expression in IRE1αKO cancer cells rescues normal tumor progression. We have developed an IRE1α gene signature that predicts immune cell infiltration and overall survival in human NSCLC. Our study unveils an immunoregulatory role for cancer cell-intrinsic IRE1α activation and suggests that targeting this pathway may help enhance anti-tumor immunity in NSCLC.

Introduction Technology has improved the diagnosis and therapeutic tracking of sleep disorders, but remains fragmented. For example, the laboratory polysomnogram and home sleep apnea test is disconnected from the data available from therapy devices. Remote patient monitoring has focused more on apnea metrics rather than sleep quality. There is some value to a device or system which can be used for diagnosis, risk stratification, and tracking of clinical outcomes across different sleep disorders. Methods A method which offers the potential of end-to-end utility in sleep medicine is the cardiopulmonary coupling (CPC) spectrogram. Originally an ECG-base device, the current form is that of a Ring oximeter (finger) with an integrated activity monitor. The oximeter data is processed via Bluetooth in a smartphone, and then transferred to a Cloud-based analysis (www.sleepimage.com). The CPC analysis integrates heart rate variability and ECG or photoplethysmography-derived respiration. Outputs include a Sleep Quality Index and percentage of stable sleep (high frequency coupling which covaries with delta power on the electroencephalogram and is associated with blood pressure dipping), unstable sleep (low frequency coupling), an apnea-hypopnea index, oxygen desaturation index, heart rate profile across the night (dipping or non-dipping heart rate), total sleep time and sleep efficiency. Data from two clinical systems were used: Empower Sleep (an Internet-based “digital sleep clinic”) and Sleep Continuum (a sleep tracking service offered by Neurocare, Inc, in the Boston area). Results Data from over 500 patients across a range of sleep disorders show end-to-end utility of the SleepImage system. These include diagnosis of sleep apnea including multi-night recording, sleep apnea phenotyping (high loop gain and sleep fragmentation), remote patient monitoring during CPAP or oral appliance use, supportive information for diagnosis of hypersomnia, tracking medication effects (sodium oxybate, sedative-hypnotics, acetazolamide), sleep-wake instability in mood disorders and Long-Covid, impaired sleep quality in chronic fatigue syndrome and restless legs syndrome, off-label therapies for central sleep apnea (oxygen, acetazolamide, buspirone), and sleep fragmentation in Long-Covid and parasomnia. Examples from each of these clinical use cases will be shown. Conclusion The CPC-based SleepImage Ring system can provide complete diagnostic, phenotyping, and tracking functions. Support (if any) The Institute for Personalized Sleep Health, BIDMC, Empower Sleep

Recognition of pathogen-associated molecular patterns can trigger the IRE1α arm of the endoplasmic reticulum (ER) stress response in immune cells. IRE1α activation has been shown to maintain ER homeostasis while simultaneously coordinating diverse immunomodulatory programs in the setting of bacterial and viral infections. However, the role of IRE1α signaling in innate immune responses to fungal pathogens is unknown. Here we report that systemic infection with the fungusCandida albicanscauses severe renal immunopathology by triggering inflammatory IRE1α hyperactivation in host myeloid cells. Mechanistically, sensing of fungal β-glucans by the C-type lectin receptor Dectin-1 induced Src–Syk–NOX-dependent accumulation of intracellular reactive oxygen species and the ensuing generation of lipid peroxidation byproducts that sustained IRE1α activation. Selective deletion of IRE1α in leukocytes, or treatment with an IRE1α pharmacological inhibitor, reduced detrimental inflammatory responses in the kidney and extended survival in mice systemically infected withC. albicans. Hence, controlling IRE1α overactivation may be useful to impede the fatal immunopathogenic progression of disseminated candidiasis.One sentence summaryInnate IRE1α signaling in disseminated candidiasis

Vast amounts of molecular data are being collected on tumor samples, which provide unique opportunities for discovering trends within and between cancer subtypes. Such cross-cancer analyses require computational methods that enable intuitive and interactive browsing of thousands of samples based on their molecular similarity. We created a portal called TumorMap to assist in exploration and statistical interrogation of high-dimensional complex “omics” data in an interactive and easily interpretable way. In the TumorMap, samples are arranged on a hexagonal grid based on their similarity to one another in the original genomic space and are rendered with Google's Map technology. While the important feature of this public portal is the ability for the users to build maps from their own data, we pre-built genomic maps from several previously published projects. We demonstrate the utility of this portal by presenting results obtained from The Cancer Genome Atlas project data. Cancer Res; 77(21); e111–4. ©2017 AACR.

Inositol-requiring enzyme 1[α] (IRE1[α])–X-box binding protein spliced (XBP1) signaling maintains endoplasmic reticulum (ER) homeostasis while controlling immunometabolic processes. Yet, the physiological consequences of IRE1a–XBP1 activation in leukocytes remain unexplored. We found that induction of prostaglandin-endoperoxide synthase 2 (Ptgs2/Cox-2) and prostaglandin E synthase (Ptges/mPGES-1) was compromised in IRE1a-deficient myeloid cells undergoing ER stress or stimulated through pattern recognition receptors. Inducible biosynthesis of prostaglandins, including the pro-algesic mediator prostaglandin E2 (PGE2), was decreased in myeloid cells that lack IRE1a or XBP1 but not other ER stress sensors. Functional XBP1 transactivated the human PTGS2 and PTGES genes to enable optimal PGE2 production. Mice that lack IRE1α–XBP1 in leukocytes, or that were treated with IRE1a inhibitors, demonstrated reduced pain behaviors in PGE2-dependent models of pain. Thus, IRE1a–XBP1 is a mediator of prostaglandin biosynthesis and a potential target to control pain.

In this paper, we present Biomedical Multi-Task Deep Neural Network (Bio-MTDNN) on the NLI task of MediQA 2019 challenge. Bio-MTDNN utilizes “transfer learning” based paradigm where not only the source and target domains are different but also the source and target tasks are varied, although related. Further, Bio-MTDNN integrates knowledge from external sources such as clinical databases (UMLS) enhancing its performance on the clinical domain. Our proposed method outperformed the official baseline and other prior models (such as ESIM and Infersent on dev set) by a considerable margin as evident from our experimental results.

To demonstrate that use of a minimally invasive catheter reduces endotracheal tube (ETT) malposition rate after intubation.This study is a multi-center, prospective observational cohort of intubated patients in the medical intensive care unit. The catheter was inserted into the ETT immediately after intubation. The ETT was adjusted accordingly based on qualitative color markers on the catheter. A confirmatory chest radiograph was obtained to determine the ETT position. Malposition of the ETT was defined by the distal ETT not being within 2-5 cm above the carina.Sixty-nine patients were enrolled, age 56.2 ± 19.5 years, body mass index 31.0 ± 13.8 kg/mWith use of an ETT positioning catheter after intubation, the ETT malposition rate was reduced by 82%. This catheter-based system was safe, and its use may perhaps decrease the need for the post-intubation chest radiograph.

Social media-based text mining in healthcare has received special attention in recent times due to the enhanced accessibility of social media sites like Twitter. The increasing trend of spreading important information in distress can help patients reach out to prospective blood donors in a time bound manner. However such manual efforts are mostly inefficient due to the limited network of a user. In a novel step to solve this problem, we present an annotated Emergency Blood Donation Request (EBDR) dataset to classify tweets referring to the necessity of urgent blood donation requirement. Additionally, we also present an automated feature-based SVM classification technique that can help selective EBDR tweets reach relevant personals as well as medical authorities. Our experiments also present a quantitative evidence that linguistic along with handcrafted heuristics can act as the most representative set of signals this task with an accuracy of 97.89%.

SummaryDendritic cells (DCs) are required to initiate and sustain T cell-dependent anti-cancer immunity. However, tumors often evade immune control by crippling normal DC function. The endoplasmic reticulum (ER) stress response factor XBP1 promotes intrinsic tumor growth directly, but whether it also regulates the host anti-tumor immune response is not known. Here we show that constitutive activation of XBP1 in tumor-associated DCs (tDCs) drives ovarian cancer (OvCa) progression by blunting anti-tumor immunity. XBP1 activation, fueled by lipid peroxidation byproducts, induced a triglyceride biosynthetic program in tDCs leading to abnormal lipid accumulation and subsequent inhibition of tDC capacity to support anti-tumor T cells. Accordingly, DC-specific XBP1 deletion or selective nanoparticle-mediated XBP1 silencing in tDCs restored their immunostimulatory activity in situ and extended survival by evoking protective type 1 anti-tumor responses. Targeting the ER stress response should concomitantly inhibit tumor growth and enhance anti-cancer immunity, thus offering a unique approach to cancer immunotherapy.

Tumours evade immune control by creating hostile microenvironments that perturb T cell metabolism and effector function 1?4 . However, it remains unclear how intra-tumoral T cells integrate and interpret metabolic stress signals. Here we report that ovarian cancer?an aggressive malignancy that is refractory to standard treatments and current immunotherapies 5?8 ?induces endoplasmic reticulum stress and activates the IRE1α?XBP1 arm of the unfolded protein response 9,10 in T cells to control their mitochondrial respiration and anti-tumour function. In T cells isolated from specimens collected from patients with ovarian cancer, upregulation of XBP1 was associated with decreased infiltration of T cells into tumours and with reduced IFNG mRNA expression. Malignant ascites fluid obtained from patients with ovarian cancer inhibited glucose uptake and caused N-linked protein glycosylation defects in T cells, which triggered IRE1α?XBP1 activation that suppressed mitochondrial activity and IFNγ production. Mechanistically, induction of XBP1 regulated the abundance of glutamine carriers and thus limited the influx of glutamine that is necessary to sustain mitochondrial respiration in T cells under glucose-deprived conditions. Restoring N-linked protein glycosylation, abrogating IRE1α?XBP1 activation or enforcing expression of glutamine transporters enhanced mitochondrial respiration in human T cells exposed to ovarian cancer ascites. XBP1-deficient T cells in the metastatic ovarian cancer milieu exhibited global transcriptional reprogramming and improved effector capacity. Accordingly, mice that bear ovarian cancer and lack XBP1 selectively in T cells demonstrate superior anti-tumour immunity, delayed malignant progression and increased overall survival. Controlling endoplasmic reticulum stress or targeting IRE1α?XBP1 signalling may help to restore the metabolic fitness and anti-tumour capacity of T cells in cancer hosts. Fil: Song, Minkyung. Weill Cornell Medicine; Estados Unidos Fil: Sandoval, Tito A.. Weill Cornell Medicine; Estados Unidos Fil: Chae, Chang-Suk. Weill Cornell Medicine; Estados Unidos Fil: Chopra, Sahil. Weill Cornell Medicine; Estados Unidos Fil: Tan, Chen. Weill Cornell Medicine; Estados Unidos Fil: Rutkowski, Melanie R.. University of Virginia; Estados Unidos Fil: Raundhal, Mahesh. Dana Farber Cancer Institute; Estados Unidos. Harvard Medical School; Estados Unidos Fil: Chaurio, Ricardo A.. H. Lee Moffitt Cancer Center & Research Institute; Estados Unidos Fil: Payne, Kyle K.. H. Lee Moffitt Cancer Center & Research Institute; Estados Unidos Fil: Konrad, Csaba. Weill Cornell Medicine; Estados Unidos Fil: Bettigole, Sarah E.. Quentis Therapeutics Inc.; Estados Unidos Fil: Shin, Hee Rae. Quentis Therapeutics Inc.; Estados Unidos Fil: Crowley, Michael J. P.. Weill Cornell Graduate School of Medical Sciences; Estados Unidos Fil: Cerliani, Juan Pablo. Consejo Nacional de Investigaciones Científicas y Técnicas. Instituto de Biología y Medicina Experimental. Fundación de Instituto de Biología y Medicina Experimental. Instituto de Biología y Medicina Experimental; Argentina Fil: Kossenkov, Andrew V.. The Wistar Institute; Estados Unidos Fil: Motorykin, Ievgen. Weill Cornell Medicine,; Estados Unidos Fil: Zhang, Sheng. Weill Cornell Medicine,; Estados Unidos Fil: Manfredi, Giovanni. Weill Cornell Medicine,; Estados Unidos Fil: Zamarin, Dmitriy. Memorial Sloan Kettering Cancer Center; Estados Unidos Fil: Holcomb, Kevin. Weill Cornell Medicine,; Estados Unidos Fil: Rodriguez, Paulo C.. H. Lee Moffitt Cancer Center & Research Institute; Estados Unidos Fil: Rabinovich, Gabriel Adrián. Consejo Nacional de Investigaciones Científicas y Técnicas. Instituto de Biología y Medicina Experimental. Fundación de Instituto de Biología y Medicina Experimental. Instituto de Biología y Medicina Experimental; Argentina. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Departamento de Química Biológica; Argentina Fil: Conejo Garcia, Jose R.. H. Lee Moffitt Cancer Center & Research Institute; Estados Unidos Fil: Glimcher, Laurie H.. Dana Farber Cancer Institute; Estados Unidos. Harvard Medical School; Estados Unidos Fil: Cubillos-Ruiz, Juan R.. Weill Graduate School Of Medical Sciences; Estados Unidos. Weill Graduate School Of Medical Sciences; Estados Unidos

Cancer is caused by accumulated changes in a cell's DNA sequence that disrupt the regulation of genetic networks. Mutation type, cell of origin, and tissue microenvironment all influence the initiation and progression of disease. The Cancer Genome Atlas (TCGA) catalogues the molecular changes of thousands of tumor samples of various tumor types using different data modalities including genomic, transcriptomic, proteomic, and epigenomic views. The goal is to find similarities amongst cancers of different tissues and to reveal clinically-relevant subtypes sharing common molecular abnormalities. While various analyses for interpreting these rich datasets exist, few methods are available to enable intuitive global overviews of these rich compendia. There is a need for methods that can tap the statistical power of large cohorts, to aid in analysis of smaller cohorts and of individual patient samples. Patterns present in many tumors may reveal driving genomic aberrations and pathway signatures that inform therapy. Here, we present a TumorMap, a tool that generates a map of cancer samples for interactive exploration, data overlay visualization, and statistical analysis. TumorMap arranges samples on a hexagonal 2-dimensional grid based on sample similarity. Different maps can be made for each distinct platform of data. Herein we demonstrate the utility of TumorMap for revealing commonalities between cancers of different tissue types, and its ability to aid in pan-cancer hypothesis generation. TumorMap maps for various cancer cohorts are available for free online at http://tumormap.ucsc.edu. Citation Format: Yulia Newton, Adam Novak, Teresa Swatlowski, Sahil Chopra, Sofie Salama, Olena Morozova, David Haussler, Joshua Stuart. UCSC TumorMap: Exploring cancer signatures on an interactive dynamic landscape. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr LB-290.

We applied biologically-motivated feature transformations coupled with established machine learning methods to predict gene essentiality in CCLE cell line models. By leveraging additional large datasets, such as The Cancer Genome Atlas PanCancer12 data and MSigDB pathway definitions, we improved the robustness and biological interpretability of our models. We developed a multi-pathway learning (MPL) approach that associates a genetic pathway from MSigDB with a distinct kernel for use in a multiple kernel learning setting. We evaluated the performance of MPL compared to several other regression methods including random forests, kernel ridge regression, and elastic net linear models, We combined multiple approaches using an ensemble technique on the diverse set of predictors. We found that the best performing method was an ensemble combining MPL and random forest predictions. Both models utilized features derived from both gene expression and copy number data, the latter of which were filtered to those predicted as driver events in prior pan-cancer studies. The ensemble method was a joint winner in the recent DREAM 9 gene essentiality prediction challenge. MPL also demonstrated merit as a feature selector when used with other downstream methods. The ensemble performed best at predicting the essentiality of genes involved in cell cycle control (cyclins and cyclin-dependent kinases), protein degradation (proteasome complex), cell proliferation signaling (sonic hedgehog, Aurora-B, RAC1), apoptosis (RB1,TP53) and hypoxia response (VEGF, VHL). Many of the key genes in those pathways are known to be drivers of cancer progression, suggesting our method's utility as a biomarker for detecting key tumorigenic events. The advantage of MPL is that mechanistically coherent gene sets are automatically selected as high scoring pathway kernels (HSPKs). We investigated whether the HSPKs identify cellular processes relevant to the loss of key genes. To do this, we inspected the HSPKs for a few of the most abundantly mutated genes in cancer. The MPL predictor for TP53 included the targets of this transcription factor as well as HSPKs involved in apoptosis, a cellular process regulated by TP53. The retinoblastoma gene (RB1) MPL predictor included RB1 targets as well as HSPKs involved in the regulation of histone deacetylase (HDAC) that interacts with RB1 to suppress DNA synthesis. These findings suggest trends in the MPL results could reveal a pathway-level view of the synthetic lethal architecture of cells. Such a map, that links patterns of pathway expression to potential genetic vulnerabilities, could provide an invaluable tool for exploring new avenues to target cancer cells. Citation Format: Vladislav Uzunangelov, Evan Paull, Sahil Chopra, Daniel Carlin, Adrian Bivol, Kyle Ellrott, Kiley Graim, Yulia Newton, Sam Ng, Artem Sokolov, Joshua Stuart. Multiple Pathway Learning accurately predicts gene essentiality in the Cancer Cell Line Encyclopedia. [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 PR10.

We applied biologically-motivated feature transformations coupled with established machine learning methods to predict gene essentiality in CCLE cell line models. By leveraging additional large datasets, such as The Cancer Genome Atlas PanCancer12 data and MSigDB pathway definitions, we improved the robustness and biological interpretability of our models. We developed a multi-pathway learning (MPL) approach that associates a genetic pathway from MSigDB with a distinct kernel for use in a multiple kernel learning setting. We evaluated the performance of MPL compared to several other regression methods including random forests, kernel ridge regression, and elastic net linear models, We combined multiple approaches using an ensemble technique on the diverse set of predictors. We found that the best performing method was an ensemble combining MPL and random forest predictions. Both models utilized features derived from both gene expression and copy number data, the latter of which were filtered to those predicted as driver events in prior pan-cancer studies. The ensemble method was a joint winner in the recent DREAM 9 gene essentiality prediction challenge. MPL also demonstrated merit as a feature selector when used with other downstream methods. The ensemble performed best at predicting the essentiality of genes involved in cell cycle control (cyclins and cyclin-dependent kinases), protein degradation (proteasome complex), cell proliferation signaling (sonic hedgehog, Aurora-B, RAC1), apoptosis (RB1,TP53) and hypoxia response (VEGF, VHL). Many of the key genes in those pathways are known to be drivers of cancer progression, suggesting our method's utility as a biomarker for detecting key tumorigenic events. The advantage of MPL is that mechanistically coherent gene sets are automatically selected as high scoring pathway kernels (HSPKs). We investigated whether the HSPKs identify cellular processes relevant to the loss of key genes. To do this, we inspected the HSPKs for a few of the most abundantly mutated genes in cancer. The MPL predictor for TP53 included the targets of this transcription factor as well as HSPKs involved in apoptosis, a cellular process regulated by TP53. The retinoblastoma gene (RB1) MPL predictor included RB1 targets as well as HSPKs involved in the regulation of histone deacetylase (HDAC) that interacts with RB1 to suppress DNA synthesis. These findings suggest trends in the MPL results could reveal a pathway-level view of the synthetic lethal architecture of cells. Such a map, that links patterns of pathway expression to potential genetic vulnerabilities, could provide an invaluable tool for exploring new avenues to target cancer cells. Citation Format: Vladislav Uzunangelov, Evan Paull, Sahil Chopra, Daniel Carlin, Adrian Bivol, Kyle Ellrott, Kiley Graim, Yulia Newton, Sam Ng, Artem Sokolov, Joshua Stuart. Multiple Pathway Learning accurately predicts gene essentiality in the Cancer Cell Line Encyclopedia. [abstract]. In: Proceedings of the AACR Special Conference on Computational and Systems Biology of Cancer; Feb 8-11 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 2):Abstract nr PR02.