Biased Deep Learning Methods in Detection of COVID-19 Using CT Images: A Challenge Mounted by Subject-Wise-Split ISFCT Dataset.

Accurate detection of respiratory system damage including COVID-19 is considered one of the crucial applications of deep learning (DL) models using CT images. However, the main shortcoming of the published works has been unreliable reported accuracy and the lack of repeatability with new datasets, mainly due to slice-wise splits of the data, creating dependency between training and test sets due to shared data across the sets. We introduce a new dataset of CT images (ISFCT Dataset) with labels indicating the subject-wise split to train and test our DL algorithms in an unbiased manner. We also use this dataset to validate the real performance of the published works in a subject-wise data split. Another key feature provides more specific labels (eight characteristic lung features) rather than being limited to COVID-19 and healthy labels. We show that the reported high accuracy of the existing models on current slice-wise splits is not repeatable for subject-wise splits, and distribution differences between data splits are demonstrated using t-distribution stochastic neighbor embedding. We indicate that, by examining subject-wise data splitting, less complicated models show competitive results compared to the exiting complicated models, demonstrating that complex models do not necessarily generate accurate and repeatable results.

A primary hierarchically organized patient-derived model enables in depth interrogation of stemness driven by the coding and non-coding genome.

Many cancers are organized as cellular hierarchies sustained by cancer stem cells (CSC), whose eradication is crucial for achieving long-term remission. Difficulties to isolate and undertake in vitro and in vivo experimental studies of rare CSC under conditions that preserve their original properties currently constitute a bottleneck for identifying molecular mechanisms involving coding and non-coding genomic regions that govern stemness. We focussed on acute myeloid leukemia (AML) as a paradigm of the CSC model and developed a patient-derived system termed OCI-AML22 that recapitulates the cellular hierarchy driven by leukemia stem cells (LSC). Through classical flow sorting and functional analyses, we established that a single phenotypic population is highly enriched for LSC. The LSC fraction can be easily isolated and serially expanded in culture or in xenografts while faithfully recapitulating functional, transcriptional and epigenetic features of primary LSCs. A novel non-coding regulatory element was identified with a new computational approach using functionally validated primary AML LSC fractions and its role in LSC stemness validated through efficient CRISPR editing using methods optimized for OCI-AML22 LSC. Collectively, OCI-AML22 constitutes a valuable resource to uncover mechanisms governing CSC driven malignancies.

Multiscale interactome analysis coupled with off-target drug predictions reveals drug repurposing candidates for human coronavirus disease.

The COVID-19 pandemic has highlighted the urgent need for the identification of new antiviral drug therapies for a variety of diseases. COVID-19 is caused by infection with the human coronavirus SARS-CoV-2, while other related human coronaviruses cause diseases ranging from severe respiratory infections to the common cold. We developed a computational approach to identify new antiviral drug targets and repurpose clinically-relevant drug compounds for the treatment of a range of human coronavirus diseases. Our approach is based on graph convolutional networks (GCN) and involves multiscale host-virus interactome analysis coupled to off-target drug predictions. Cell-based experimental assessment reveals several clinically-relevant drug repurposing candidates predicted by the in silico analyses to have antiviral activity against human coronavirus infection. In particular, we identify the MET inhibitor capmatinib as having potent and broad antiviral activity against several coronaviruses in a MET-independent manner, as well as novel roles for host cell proteins such as IRAK1/4 in supporting human coronavirus infection, which can inform further drug discovery studies.

Biological and therapeutic implications of a unique subtype of NPM1 mutated AML.

In acute myeloid leukemia (AML), molecular heterogeneity across patients constitutes a major challenge for prognosis and therapy. AML with NPM1 mutation is a distinct genetic entity in the revised World Health Organization classification. However, differing patterns of co-mutation and response to therapy within this group necessitate further stratification. Here we report two distinct subtypes within NPM1 mutated AML patients, which we label as primitive and committed based on the respective presence or absence of a stem cell signature. Using gene expression (RNA-seq), epigenomic (ATAC-seq) and immunophenotyping (CyToF) analysis, we associate each subtype with specific molecular characteristics, disease differentiation state and patient survival. Using ex vivo drug sensitivity profiling, we show a differential drug response of the subtypes to specific kinase inhibitors, irrespective of the FLT3-ITD status. Differential drug responses of the primitive and committed subtype are validated in an independent AML cohort. Our results highlight heterogeneity among NPM1 mutated AML patient samples based on stemness and suggest that the addition of kinase inhibitors to the treatment of cases with the primitive signature, lacking FLT3-ITD, could have therapeutic benefit.

Large organized chromatin lysine domains help distinguish primitive from differentiated cell populations.

The human genome is partitioned into a collection of genomic features, inclusive of genes, transposable elements, lamina interacting regions, early replicating control elements and cis-regulatory elements, such as promoters, enhancers, and anchors of chromatin interactions. Uneven distribution of these features within chromosomes gives rise to clusters, such as topologically associating domains (TADs), lamina-associated domains, clusters of cis-regulatory elements or large organized chromatin lysine (K) domains (LOCKs). Here we show that LOCKs from diverse histone modifications discriminate primitive from differentiated cell types. Active LOCKs (H3K4me1, H3K4me3 and H3K27ac) cover a higher fraction of the genome in primitive compared to differentiated cell types while repressive LOCKs (H3K9me3, H3K27me3 and H3K36me3) do not. Active LOCKs in differentiated cells lie proximal to highly expressed genes while active LOCKs in primitive cells tend to be bivalent. Genes proximal to bivalent LOCKs are minimally expressed in primitive cells. Furthermore, bivalent LOCKs populate TAD boundaries and are preferentially bound by regulators of chromatin interactions, including CTCF, RAD21 and ZNF143. Together, our results argue that LOCKs discriminate primitive from differentiated cell populations.

Single-cell chromatin accessibility profiling of glioblastoma identifies an invasive cancer stem cell population associated with lower survival.

Chromatin accessibility discriminates stem from mature cell populations, enabling the identification of primitive stem-like cells in primary tumors, such as glioblastoma (GBM) where self-renewing cells driving cancer progression and recurrence are prime targets for therapeutic intervention. We show, using single-cell chromatin accessibility, that primary human GBMs harbor a heterogeneous self-renewing population whose diversity is captured in patient-derived glioblastoma stem cells (GSCs). In-depth characterization of chromatin accessibility in GSCs identifies three GSC states: Reactive, Constructive, and Invasive, each governed by uniquely essential transcription factors and present within GBMs in varying proportions. Orthotopic xenografts reveal that GSC states associate with survival, and identify an invasive GSC signature predictive of low patient survival, in line with the higher invasive properties of Invasive state GSCs compared to Reactive and Constructive GSCs as shown by in vitro and in vivo assays. Our chromatin-driven characterization of GSC states improves prognostic precision and identifies dependencies to guide combination therapies.

Pathway-Based Drug Response Prediction Using Similarity Identification in Gene Expression.

Lapatinib and trastuzumab (Herceptin) are targeted therapies designed for patients with HER2+ breast tumors. Although these therapies improved survival rates of patients with this tumor type, not all the patients harboring HER2 amplification respond to these drugs. The NeoALTTO clinical trial was designed to test whether a higher response rate can be achieved by combining lapatinib and trastuzumab. Although the combination therapy showed almost double the response rate compared to the monotherapies, 40% of the patients did not respond to the treatment. In this study, we sought to identify biomarkers of HER2+ breast cancer patients' response to drugs relying on gene expression profiles of tumors. We show that univariate gene expression-based biomarkers are significant but weak predictors of drug response. We further show that pathway activities, estimated from gene expression patterns quantified using the recent transcriptional similarity coefficient (TSC) between the tumor samples, yield high predictive value for therapy response (concordance index >0.8, p < 0.05). Moreover, machine learning models, built using multiple algorithms including logistic regression, naive Bayes, random forest, k-nearest neighbor, and support vector machine, for predicting drug response in the NeoALTTO clinical trial, resulted in lower performance compared to our pathway-based approach. Our results indicate that transcriptional similarity of biological pathways can be used to predict lapatinib and trastuzumab response in HER2+ breast cancer.

SYNERGxDB: an integrative pharmacogenomic portal to identify synergistic drug combinations for precision oncology.

Drug-combination data portals have recently been introduced to mine huge amounts of pharmacological data with the aim of improving current chemotherapy strategies. However, these portals have only been investigated for isolated datasets, and molecular profiles of cancer cell lines are lacking. Here we developed a cloud-based pharmacogenomics portal called SYNERGxDB (http://SYNERGxDB.ca/) that integrates multiple high-throughput drug-combination studies with molecular and pharmacological profiles of a large panel of cancer cell lines. This portal enables the identification of synergistic drug combinations through harmonization and unified computational analysis. We integrated nine of the largest drug combination datasets from both academic groups and pharmaceutical companies, resulting in 22 507 unique drug combinations (1977 unique compounds) screened against 151 cancer cell lines. This data compendium includes metabolomics, gene expression, copy number and mutation profiles of the cancer cell lines. In addition, SYNERGxDB provides analytical tools to discover effective therapeutic combinations and predictive biomarkers across cancer, including specific types. Combining molecular and pharmacological profiles, we systematically explored the large space of univariate predictors of drug synergism. SYNERGxDB constitutes a comprehensive resource that opens new avenues of research for exploring the mechanism of action for drug synergy with the potential of identifying new treatment strategies for cancer patients.

Noncoding mutations target cis-regulatory elements of the FOXA1 plexus in prostate cancer.

Prostate cancer is the second most commonly diagnosed malignancy among men worldwide. Recurrently mutated in primary and metastatic prostate tumors, FOXA1 encodes a pioneer transcription factor involved in disease onset and progression through both androgen receptor-dependent and androgen receptor-independent mechanisms. Despite its oncogenic properties however, the regulation of FOXA1 expression remains unknown. Here, we identify a set of six cis-regulatory elements in the FOXA1 regulatory plexus harboring somatic single-nucleotide variants in primary prostate tumors. We find that deletion and repression of these cis-regulatory elements significantly decreases FOXA1 expression and prostate cancer cell growth. Six of the ten single-nucleotide variants mapping to FOXA1 regulatory plexus significantly alter the transactivation potential of cis-regulatory elements by modulating the binding of transcription factors. Collectively, our results identify cis-regulatory elements within the FOXA1 plexus mutated in primary prostate tumors as potential targets for therapeutic intervention.

Identifying clusters of cis-regulatory elements underpinning TAD structures and lineage-specific regulatory networks.

Cellular identity relies on cell-type-specific gene expression controlled at the transcriptional level by cis-regulatory elements (CREs). CREs are unevenly distributed across the genome, giving rise to individual CREs and clusters of CREs (COREs). Technical and biological features hinder CORE identification. We addressed these issues by developing an unsupervised machine learning approach termed clustering of genomic regions analysis method (CREAM). CREAM automates CORE detection from chromatin accessibility profiles that are enriched in CREs strongly bound by master transcription regulators, proximal to highly expressed and essential genes, and discriminating cell identity. Although COREs share similarities with super-enhancers, we highlight differences in terms of the genomic distribution and structure of these cis-regulatory units. We further show the enhanced value of COREs over super-enhancers to identify master transcription regulators, highly expressed and essential genes defining cell identity. COREs enrich at topologically associated domain (TAD) boundaries. They are also preferentially bound by the chromatin looping factors CTCF and cohesin, in contrast to super-enhancers, forming clusters of CTCF and cohesin binding regions and defining homotypic clusters of transcription regulator binding regions (HCTs). Finally, we show the clinical utility of CREAM to identify COREs across chromatin accessibility profiles to stratify more than 400 tumor samples according to their cancer type and to delineate cancer type-specific active biological pathways. Collectively, our results support the utility of CREAM to delineate COREs underlying, with greater accuracy than individual CREs or super-enhancers, the cell-type-specific biological underpinning across a wide range of normal and cancer cell types.

SIGN: similarity identification in gene expression.

MOTIVATION: High-throughput molecular profiles of human cells have been used in predictive computational approaches for stratification of healthy and malignant phenotypes and identification of their biological states. In this regard, pathway activities have been used as biological features in unsupervised and supervised learning schemes. RESULTS: We developed SIGN (Similarity Identification in Gene expressioN), a flexible open-source R package facilitating the use of pathway activities and their expression patterns to identify similarities between biological samples. We defined a new measure, the transcriptional similarity coefficient, which captures similarity of gene expression patterns, instead of quantifying overall activity, in biological pathways between the samples. To demonstrate the utility of SIGN in biomedical research, we establish that SIGN discriminates subtypes of breast tumors and patients with good or poor overall survival. SIGN outperforms the best models in DREAM challenge in predicting survival of breast cancer patients using the data from the Molecular Taxonomy of Breast Cancer International Consortium. In summary, SIGN can be used as a new tool for interrogating pathway activity and gene expression patterns in unsupervised and supervised learning schemes to improve prognostic risk estimation for cancer patients by the biomedical research community. AVAILABILITY AND IMPLEMENTATION: An open-source R package is available (https://cran.r-project.org/web/packages/SIGN/).

Predictive approaches for drug combination discovery in cancer.

Drug combinations have been proposed as a promising therapeutic strategy to overcome drug resistance and improve efficacy of monotherapy regimens in cancer. This strategy aims at targeting multiple components of this complex disease. Despite the increasing number of drug combinations in use, many of them were empirically found in the clinic, and the molecular mechanisms underlying these drug combinations are often unclear. These challenges call for rational, systematic approaches for drug combination discovery. Although high-throughput screening of single-agent therapeutics has been successfully implemented, it is not feasible to test all possible drug combinations, even for a reduced subset of anticancer drugs. Hence, in vitro and in vivo screening of a large number of drug combinations are not practical. Therefore, devising computational methods to efficiently explore the space of drug combinations and to discover efficacious combinations has attracted a lot of attention from the scientific community in the past few years. Nevertheless, in the absence of consensus regarding the computational approaches used to predict efficacious drug combinations, a plethora of methods, techniques and hypotheses have been developed to date, while the research field lacks an elaborate categorization of the existing computational methods and the available data sources. In this manuscript, we review and categorize the state-of-the-art computational approaches for drug combination prediction, and elaborate on the limitations of these methods and the existing challenges. We also discuss about the recent pan-cancer drug combination data sets and their importance in revising the available methods or developing more performant approaches.

Tissue specificity of in vitro drug sensitivity.

Objectives: We sought to investigate the tissue specificity of drug sensitivities in large-scale pharmacological studies and compare these associations to those found in drug clinical indications. Materials and Methods: We leveraged the curated cell line response data from PharmacoGx and applied an enrichment algorithm on drug sensitivity values' area under the drug dose-response curves (AUCs) with and without adjustment for general level of drug sensitivity. Results: We observed tissue specificity in 63% of tested drugs, with 8% of total interactions deemed significant (false discovery rate <0.05). By restricting the drug-tissue interactions to those with AUC > 0.2, we found that in 52% of interactions, the tissue was predictive of drug sensitivity (concordance index > 0.65). When compared with clinical indications, the observed overlap was weak (Matthew correlation coefficient, MCC = 0.0003, P > .10). Discussion: While drugs exhibit significant tissue specificity in vitro, there is little overlap with clinical indications. This can be attributed to factors such as underlying biological differences between in vitro models and patient tumors, or the inability of tissue-specific drugs to bring additional benefits beyond gold standard treatments during clinical trials. Conclusion: Our meta-analysis of pan-cancer drug screening datasets indicates that most tested drugs exhibit tissue-specific sensitivities in a large panel of cancer cell lines. However, the observed preclinical results do not translate to the clinical setting. Our results suggest that additional research into showing parallels between preclinical and clinical data is required to increase the translational potential of in vitro drug screening.

Mathematical modelling of phenotypic plasticity and conversion to a stem-cell state under hypoxia.

Hypoxia, or oxygen deficiency, is known to be associated with breast tumour progression, resistance to conventional therapies and poor clinical prognosis. The epithelial-mesenchymal transition (EMT) is a process that confers invasive and migratory capabilities as well as stem cell properties to carcinoma cells thus promoting metastatic progression. In this work, we examined the impact of hypoxia on EMT-associated cancer stem cell (CSC) properties, by culturing transformed human mammary epithelial cells under normoxic and hypoxic conditions, and applying in silico mathematical modelling to simulate the impact of hypoxia on the acquisition of CSC attributes and the transitions between differentiated and stem-like states. Our results indicate that both the heterogeneity and the plasticity of the transformed cell population are enhanced by exposure to hypoxia, resulting in a shift towards a more stem-like population with increased EMT features. Our findings are further reinforced by gene expression analyses demonstrating the upregulation of EMT-related genes, as well as genes associated with therapy resistance, in hypoxic cells compared to normoxic counterparts. In conclusion, we demonstrate that mathematical modelling can be used to simulate the role of hypoxia as a key contributor to the plasticity and heterogeneity of transformed human mammary epithelial cells.