In line with recent advances in neural drug design and sensitivity prediction, we propose a novel architecture for interpretable prediction of anticancer compound sensitivity using a multimodal attention-based convolutional encoder. Our model is based on the three key pillars of drug sensitivity: compounds’ structure in the form of a SMILES sequence, gene expression profiles of tumors and prior knowledge on intracellular interactions from protein-protein interaction networks. We demonstrate that our multiscale convolutional attention-based encoder significantly outperforms a baseline model trained on Morgan fingerprints, a selection of encoders based on SMILES as well as previously reported state of the art for multimodal drug sensitivity prediction (R2 = 0.86 and RMSE = 0.89). Moreover, the explainability of our approach is demonstrated by a thorough analysis of the attention weights. We show that the attended genes significantly enrich apoptotic processes and that the drug attention is strongly correlated with a standard chemical structure similarity index. Finally, we report a case study of two receptor tyrosine kinase (RTK) inhibitors acting on a leukemia cell line, showcasing the ability of the model to focus on informative genes and submolecular regions of the two compounds. The demonstrated generalizability and the interpretability of our model testify its potential for in-silico prediction of anticancer compound efficacy on unseen cancer cells, positioning it as a valid solution for the development of personalized therapies as well as for the evaluation of candidate compounds in de novo drug design.
Prostate cancer is the second most occurring cancer in men worldwide. To better understand the mechanisms of tumorigenesis and possible treatment responses, we developed a mathematical model of prostate cancer which considers the major signalling pathways known to be deregulated. We personalised this Boolean model to molecular data to reflect the heterogeneity and specific response to perturbations of cancer patients. A total of 488 prostate samples were used to build patient-specific models and compared to available clinical data. Additionally, eight prostate cell line-specific models were built to validate our approach with dose-response data of several drugs. The effects of single and combined drugs were tested in these models under different growth conditions. We identified 15 actionable points of interventions in one cell line-specific model whose inactivation hinders tumorigenesis. To validate these results, we tested nine small molecule inhibitors of five of those putative targets and found a dose-dependent effect on four of them, notably those targeting HSP90 and PI3K. These results highlight the predictive power of our personalised Boolean models and illustrate how they can be used for precision oncology.
Proteochemometric (PCM) modelling is a computational method to model the bioactivity of multiple ligands against multiple related protein targets simultaneously. Hence it has been found to be particularly useful when exploring the selectivity and promiscuity of ligands on different proteins. In this review, we will firstly provide a brief introduction to the main concepts of PCM for readers new to the field. The next part focuses on recent technical advances, including the application of support vector machines (SVMs) using different kernel functions, random forests, Gaussian processes and collaborative filtering. The subsequent section will then describe some novel practical applications of PCM in the medicinal chemistry field, including studies on GPCRs, kinases, viral proteins (e.g. from HIV) and epigenetic targets such as histone deacetylases. Finally, we will conclude by summarizing novel developments in PCM, which we expect to gain further importance in the future. These developments include adding three-dimensional protein target information, application of PCM to the prediction of binding energies, and application of the concept in the fields of pharmacogenomics and toxicogenomics. This review is an update to a related publication in 2011 and it mainly focuses on developments in the field since then.
Transcriptional perturbation signatures are valuable data sources for functional genomics. Linking perturbation signatures to screenings opens the possibility to model cellular phenotypes from expression data and to identify efficacious drugs. We linked perturbation transcriptomics data from the LINCS-L1000 project with cell viability information upon genetic (Achilles project) and chemical (CTRP screen) perturbations yielding more than 90 000 signature–viability pairs. An integrated analysis showed that the cell viability signature is a major factor underlying perturbation signatures. The signature is linked to transcription factors regulating cell death, proliferation and division time. We used the cell viability–signature relationship to predict viability from transcriptomics signatures, and identified and validated compounds that induce cell death in tumor cell lines. We showed that cellular toxicity can lead to unexpected similarity of signatures, confounding mechanism of action discovery. Consensus compound signatures predicted cell-specific drug sensitivity, even if the signature is not measured in the same cell line, and outperformed conventional drug-specific features. Our results can help in understanding mechanisms behind cell death and removing confounding factors of transcriptomic perturbation screens. To interactively browse our results and predict cell viability in new gene expression samples, we developed CEVIChE (CEll VIability Calculator from gene Expression; https://saezlab.shinyapps.io/ceviche/).
25 October 2013 in Journal of Chemical Information and Modeling
Accurate methods to predict solubility from molecular structure are highly sought after in the chemical sciences. To assess the state of the art, the American Chemical Society organized a "Second Solubility Challenge" in 2019, in which competitors were invited to submit blinded predictions of the solubilities of 132 drug-like molecules. In the first part of this article, we describe the development of two models that were submitted to the Blind Challenge in 2019 but which have not previously been reported. These models were based on computationally inexpensive molecular descriptors and traditional machine learning algorithms and were trained on a relatively small data set of 300 molecules. In the second part of the article, to test the hypothesis that predictions would improve with more advanced algorithms and higher volumes of training data, we compare these original predictions with those made after the deadline using deep learning models trained on larger solubility data sets consisting of 2999 and 5697 molecules. The results show that there are several algorithms that are able to obtain near state-of-the-art performance on the solubility challenge data sets, with the best model, a graph convolutional neural network, resulting in an RMSE of 0.86 log units. Critical analysis of the models reveals systematic differences between the performance of models using certain feature sets and training data sets. The results suggest that careful selection of high quality training data from relevant regions of chemical space is critical for prediction accuracy but that other methodological issues remain problematic for machine learning solubility models, such as the difficulty in modeling complex chemical spaces from sparse training data sets.