published version features the final layout of the paper including the volume, issue and page numbers. Link to publication General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the "Taverne" license above, please follow below link for the End User

Cancer is a leading cause of death worldwide. Identifying the best treatment using computational models to personalize drug response prediction holds great promise to improve patient’s chances of successful recovery. Unfortunately, the computational task of predicting drug response is very challenging, partially due to the limitations of the available data and partially due to algorithmic shortcomings. The recent advances in deep learning may open a new chapter in the search for computational drug response prediction models and ultimately result in more accurate tools for therapy response. This review provides an overview of the computational challenges and advances in drug response prediction, and focuses on comparing the machine learning techniques to be of utmost practical use for clinicians and machine learning non-experts. The incorporation of new data modalities such as single-cell profiling, along with techniques that rapidly find effective drug combinations will likely be instrumental in improving cancer care.

Recent cancer pharmacogenomic studies profiled large panels of cell lines against hundreds of approved drugs and experimental chemical compounds. The overarching goal of these screens is to measure sensitivity of cell lines to chemical perturbations, correlate these measures to genomic features, and thereby develop novel predictors of drug response. However, leveraging these valuable data is challenging due to the lack of standards for annotating cell lines and chemical compounds, and quantifying drug response. Moreover, it has been recently shown that the complexity and complementarity of the experimental protocols used in the field result in high levels of technical and biological variation in the in vitro pharmacological profiles. There is therefore a need for new tools to facilitate rigorous comparison and integrative analysis of large-scale drug screening datasets. To address this issue, we have developed PharmacoDB (pharmacodb.pmgenomics.ca), a database integrating the largest cancer pharmacogenomic studies published to date. Here, we describe how the curation of cell line and chemical compound identifiers maximizes the overlap between datasets and how users can leverage such data to compare and extract robust drug phenotypes. PharmacoDB provides a unique resource to mine a compendium of curated cancer pharmacogenomic datasets that are otherwise disparate and difficult to integrate.

No abstract

Despite the success of machine learning applications in science, industry, and society in general, many approaches are known to be non-robust, often relying on spurious correlations to make predictions. Spuriousness occurs when some features correlate with labels but are not causal; relying on such features prevents models from generalizing to unseen environments where such correlations break. In this work, we focus on image classification and propose two data generation processes to reduce spuriousness. Given human annotations of the subset of the features responsible (causal) for the labels (e.g. bounding boxes), we modify this causal set to generate a surrogate image that no longer has the same label (i.e. a counterfactual image). We also alter non-causal features to generate images still recognized as the original labels, which helps to learn a model invariant to these features. In several challenging datasets, our data generations outperform state-of-the-art methods in accuracy when spurious correlations break, and increase the saliency focus on causal features providing better explanations.

Recent pharmacogenomic studies profiled large panels of cancer cell lines against hundreds of approved drugs and experimental chemical compounds. The overarching goal of these screens is to measure sensitivity of cell lines to chemical perturbation, correlate these measures to genomic features, and thereby develop novel predictors of drug response. However, leveraging this valuable data is challenging due to the lack of standards for annotating cell lines and chemical compounds, and quantifying drug response. Moreover, it has been recently shown that the complexity and complementarity of the experimental protocols used in the field result in high levels of technical and biological variation in the in vitro pharmacological profiles. There is therefore a need for new tools to facilitate rigorous comparison and integrative analysis of largescale drug screening datasets. To address this issue, we have developed PharmacoDB (pharmacodb.pmgenomics.ca), a database integrating the largest pharmacogenomic studies published to date. Here, we describe how the curation of cell line and chemical compound identifiers maximizes the overlap between datasets and how users can leverage such data to compare and extract robust drug phenotypes. PharmacoDB provides a unique resource to mine a compendium of curated pharmacogenomic datasets that are otherwise disparate and difficult to integrate.

<p>The supplementary file discusses the limitations of existing tools, performance of CrosstalkNet, along with more biological case studies.</p>

<div>Abstract<p>Variations in physiological conditions can rewire molecular interactions between biological compartments, which can yield novel insights into gain or loss of interactions specific to perturbations of interest. Networks are a promising tool to elucidate intercellular interactions, yet exploration of these large-scale networks remains a challenge due to their high dimensionality. To retrieve and mine interactions, we developed CrosstalkNet, a user friendly, web-based network visualization tool that provides a statistical framework to infer condition-specific interactions coupled with a community detection algorithm for bipartite graphs to identify significantly dense subnetworks. As a case study, we used CrosstalkNet to mine a set of 54 and 22 gene-expression profiles from breast tumor and normal samples, respectively, with epithelial and stromal compartments extracted via laser microdissection. We show how CrosstalkNet can be used to explore large-scale co-expression networks and to obtain insights into the biological processes that govern cross-talk between different tumor compartments.</p><p><b>Significance:</b> This web application enables researchers to mine complex networks and to decipher novel biological processes in tumor epithelial-stroma cross-talk as well as in other studies of intercompartmental interactions. <i>Cancer Res; 78(8); 2140–3. ©2018 AACR</i>.</p></div>