Prediction of Adverse Drug Reaction Linked to Protein Targets Using Network-Based Information and Machine Learning.

Drug discovery attrition rates, particularly at advanced clinical trial stages, are high because of unexpected adverse drug reactions (ADR) elicited by novel drug candidates. Predicting undesirable ADRs produced by the modulation of certain protein targets would contribute to developing safer drugs, thereby reducing economic losses associated with high attrition rates. As opposed to the more traditional drug-centric approach, we propose a target-centric approach to predict associations between protein targets and ADRs. The implementation of the predictor is based on a machine learning classifier that integrates a set of eight independent network-based features. These include a network diffusion-based score, identification of protein modules based on network clustering algorithms, functional similarity among proteins, network distance to proteins that are part of safety panels used in preclinical drug development, set of network descriptors in the form of degree and betweenness centrality measurements, and conservation. This diverse set of descriptors were used to generate predictors based on different machine learning classifiers ranging from specific models for individual ADR to higher levels of abstraction as per MEDDRA hierarchy such as system organ class. The results obtained from the different machine-learning classifiers, namely, support vector machine, random forest, and neural network were further analyzed as a meta-predictor exploiting three different voting systems, namely, jury vote, consensus vote, and red flag, obtaining different models for each of the ADRs in analysis. The level of accuracy of the predictors justifies the identification of problematic protein targets both at the level of individual ADR as well as a set of related ADRs grouped in common system organ classes. As an example, the prediction of ventricular tachycardia achieved an accuracy and precision of 0.83 and 0.90, respectively, and a Matthew correlation coefficient of 0.70. We believe that this approach is a good complement to the existing methodologies devised to foresee potential liabilities in preclinical drug discovery. The method is available through the DocTOR utility at GitHub (https://github.com/cristian931/DocTOR).

SPServer: split-statistical potentials for the analysis of protein structures and protein-protein interactions.

BACKGROUND: Statistical potentials, also named knowledge-based potentials, are scoring functions derived from empirical data that can be used to evaluate the quality of protein folds and protein-protein interaction (PPI) structures. In previous works we decomposed the statistical potentials in different terms, named Split-Statistical Potentials, accounting for the type of amino acid pairs, their hydrophobicity, solvent accessibility and type of secondary structure. These potentials have been successfully used to identify near-native structures in protein structure prediction, rank protein docking poses, and predict PPI binding affinities. RESULTS: Here, we present the SPServer, a web server that applies the Split-Statistical Potentials to analyze protein folds and protein interfaces. SPServer provides global scores as well as residue/residue-pair profiles presented as score plots and maps. This level of detail allows users to: (1) identify potentially problematic regions on protein structures; (2) identify disrupting amino acid pairs in protein interfaces; and (3) compare and analyze the quality of tertiary and quaternary structural models. CONCLUSIONS: While there are many web servers that provide scoring functions to assess the quality of either protein folds or PPI structures, SPServer integrates both aspects in a unique easy-to-use web server. Moreover, the server permits to locally assess the quality of the structures and interfaces at a residue level and provides tools to compare the local assessment between structures. SERVER ADDRESS: https://sbi.upf.edu/spserver/ .

An ensemble learning approach for modeling the systems biology of drug-induced injury.

BACKGROUND: Drug-induced liver injury (DILI) is an adverse reaction caused by the intake of drugs of common use that produces liver damage. The impact of DILI is estimated to affect around 20 in 100,000 inhabitants worldwide each year. Despite being one of the main causes of liver failure, the pathophysiology and mechanisms of DILI are poorly understood. In the present study, we developed an ensemble learning approach based on different features (CMap gene expression, chemical structures, drug targets) to predict drugs that might cause DILI and gain a better understanding of the mechanisms linked to the adverse reaction. RESULTS: We searched for gene signatures in CMap gene expression data by using two approaches: phenotype-gene associations data from DisGeNET, and a non-parametric test comparing gene expression of DILI-Concern and No-DILI-Concern drugs (as per DILIrank definitions). The average accuracy of the classifiers in both approaches was 69%. We used chemical structures as features, obtaining an accuracy of 65%. The combination of both types of features produced an accuracy around 63%, but improved the independent hold-out test up to 67%. The use of drug-target associations as feature obtained the best accuracy (70%) in the independent hold-out test. CONCLUSIONS: When using CMap gene expression data, searching for a specific gene signature among the landmark genes improves the quality of the classifiers, but it is still limited by the intrinsic noise of the dataset. When using chemical structures as a feature, the structural diversity of the known DILI-causing drugs hampers the prediction, which is a similar problem as for the use of gene expression information. The combination of both features did not improve the quality of the classifiers but increased the robustness as shown on independent hold-out tests. The use of drug-target associations as feature improved the prediction, specially the specificity, and the results were comparable to previous research studies.

Galaxy InteractoMIX: An Integrated Computational Platform for the Study of Protein-Protein Interaction Data.

Protein interactions play a crucial role among the different functions of a cell and are central to our understanding of cellular processes both in health and disease. Here we present Galaxy InteractoMIX (http://galaxy.interactomix.com), a platform composed of 13 different computational tools each addressing specific aspects of the study of protein-protein interactions, ranging from large-scale cross-species protein-wide interactomes to atomic resolution level of protein complexes. Galaxy InteractoMIX provides an intuitive interface where users can retrieve consolidated interactomics data distributed across several databases or uncover links between diseases and genes by analyzing the interactomes underlying these diseases. The platform makes possible large-scale prediction and curation protein interactions using the conservation of motifs, interology, or presence or absence of key sequence signatures. The range of structure-based tools includes modeling and analysis of protein complexes, delineation of interfaces and the modeling of peptides acting as inhibitors of protein-protein interactions. Galaxy InteractoMIX includes a range of ready-to-use workflows to run complex analyses requiring minimal intervention by users. The potential range of applications of the platform covers different aspects of life science, biomedicine, biotechnology and drug discovery where protein associations are studied.

In-silico drug repurposing study predicts the combination of pirfenidone and melatonin as a promising candidate therapy to reduce SARS-CoV-2 infection progression and respiratory distress caused by cytokine storm.

From January 2020, COVID-19 is spreading around the world producing serious respiratory symptoms in infected patients that in some cases can be complicated by the severe acute respiratory syndrome, sepsis and septic shock, multiorgan failure, including acute kidney injury and cardiac injury. Cost and time efficient approaches to reduce the burthen of the disease are needed. To find potential COVID-19 treatments among the whole arsenal of existing drugs, we combined system biology and artificial intelligence-based approaches. The drug combination of pirfenidone and melatonin has been identified as a candidate treatment that may contribute to reduce the virus infection. Starting from different drug targets the effect of the drugs converges on human proteins with a known role in SARS-CoV-2 infection cycle. Simultaneously, GUILDify v2.0 web server has been used as an alternative method to corroborate the effect of pirfenidone and melatonin against the infection of SARS-CoV-2. We have also predicted a potential therapeutic effect of the drug combination over the respiratory associated pathology, thus tackling at the same time two important issues in COVID-19. These evidences, together with the fact that from a medical point of view both drugs are considered safe and can be combined with the current standard of care treatments for COVID-19 makes this combination very attractive for treating patients at stage II, non-severe symptomatic patients with the presence of virus and those patients who are at risk of developing severe pulmonary complications.

Using collections of structural models to predict changes of binding affinity caused by mutations in protein-protein interactions.

Protein-protein interactions (PPIs) in all the molecular aspects that take place both inside and outside cells. However, determining experimentally the structure and affinity of PPIs is expensive and time consuming. Therefore, the development of computational tools, as a complement to experimental methods, is fundamental. Here, we present a computational suite: MODPIN, to model and predict the changes of binding affinity of PPIs. In this approach we use homology modeling to derive the structures of PPIs and score them using state-of-the-art scoring functions. We explore the conformational space of PPIs by generating not a single structural model but a collection of structural models with different conformations based on several templates. We apply the approach to predict the changes in free energy upon mutations and splicing variants of large datasets of PPIs to statistically quantify the quality and accuracy of the predictions. As an example, we use MODPIN to study the effect of mutations in the interaction between colicin endonuclease 9 and colicin endonuclease 2 immune protein from Escherichia coli. Finally, we have compared our results with other state-of-art methods.

In-silico simulated prototype-patients using TPMS technology to study a potential adverse effect of sacubitril and valsartan.

Unveiling the mechanism of action of a drug is key to understand the benefits and adverse reactions of a medication in an organism. However, in complex diseases such as heart diseases there is not a unique mechanism of action but a wide range of different responses depending on the patient. Exploring this collection of mechanisms is one of the clues for a future personalized medicine. The Therapeutic Performance Mapping System (TPMS) is a Systems Biology approach that generates multiple models of the mechanism of action of a drug. Each molecular mechanism generated could be associated to particular individuals, here defined as prototype-patients, hence the generation of models using TPMS technology may be used for detecting adverse effects to specific patients. TPMS operates by (1) modelling the responses in humans with an accurate description of a protein network and (2) applying a Multilayer Perceptron-like and sampling strategy to find all plausible solutions. In the present study, TPMS is applied to explore the diversity of mechanisms of action of the drug combination sacubitril/valsartan. We use TPMS to generate a wide range of models explaining the relationship between sacubitril/valsartan and heart failure (the indication), as well as evaluating their association with macular degeneration (a potential adverse effect). Among the models generated, we identify a set of mechanisms of action associated to a better response in terms of heart failure treatment, which could also be associated to macular degeneration development. Finally, a set of 30 potential biomarkers are proposed to identify mechanisms (or prototype-patients) more prone of suffering macular degeneration when presenting good heart failure response. All prototype-patients models generated are completely theoretical and therefore they do not necessarily involve clinical effects in real patients. Data and accession to software are available at http://sbi.upf.edu/data/tpms/.

GUILDify v2.0: A Tool to Identify Molecular Networks Underlying Human Diseases, Their Comorbidities and Their Druggable Targets.

The genetic basis of complex diseases involves alterations on multiple genes. Unraveling the interplay between these genetic factors is key to the discovery of new biomarkers and treatments. In 2014, we introduced GUILDify, a web server that searches for genes associated to diseases, finds novel disease genes applying various network-based prioritization algorithms and proposes candidate drugs. Here, we present GUILDify v2.0, a major update and improvement of the original method, where we have included protein interaction data for seven species and 22 human tissues and incorporated the disease-gene associations from DisGeNET. To infer potential disease relationships associated with multi-morbidities, we introduced a novel feature for estimating the genetic and functional overlap of two diseases using the top-ranking genes and the associated enrichment of biological functions and pathways (as defined by GO and Reactome). The analysis of this overlap helps to identify the mechanistic role of genes and protein-protein interactions in comorbidities. Finally, we provided an R package, guildifyR, to facilitate programmatic access to GUILDify v2.0 (http://sbi.upf.edu/guildify2).

Network, Transcriptomic and Genomic Features Differentiate Genes Relevant for Drug Response.

Understanding the mechanisms underlying drug therapeutic action and toxicity is crucial for the prevention and management of drug adverse reactions, and paves the way for a more efficient and rational drug design. The characterization of drug targets, drug metabolism proteins, and proteins associated to side effects according to their expression patterns, their tolerance to genomic variation and their role in cellular networks, is a necessary step in this direction. In this contribution, we hypothesize that different classes of proteins involved in the therapeutic effect of drugs and in their adverse effects have distinctive transcriptomics, genomics and network features. We explored the properties of these proteins within global and organ-specific interactomes, using multi-scale network features, evaluated their gene expression profiles in different organs and tissues, and assessed their tolerance to loss-of-function variants leveraging data from 60K subjects. We found that drug targets that mediate side effects are more central in cellular networks, more intolerant to loss-of-function variation, and show a wider breadth of tissue expression than targets not mediating side effects. In contrast, drug metabolizing enzymes and transporters are less central in the interactome, more tolerant to deleterious variants, and are more constrained in their tissue expression pattern. Our findings highlight distinctive features of proteins related to drug action, which could be applied to prioritize drugs with fewer probabilities of causing side effects.

Proximal Pathway Enrichment Analysis for Targeting Comorbid Diseases via Network Endopharmacology.

The past decades have witnessed a paradigm shift from the traditional drug discovery shaped around the idea of “one target, one disease” to polypharmacology (multiple targets, one disease). Given the lack of clear-cut boundaries across disease (endo)phenotypes and genetic heterogeneity across patients, a natural extension to the current polypharmacology paradigm is to target common biological pathways involved in diseases via endopharmacology (multiple targets, multiple diseases). In this study, we present proximal pathway enrichment analysis (PxEA) for pinpointing drugs that target common disease pathways towards network endopharmacology. PxEA uses the topology information of the network of interactions between disease genes, pathway genes, drug targets and other proteins to rank drugs by their interactome-based proximity to pathways shared across multiple diseases, providing unprecedented drug repurposing opportunities. Using PxEA, we show that many drugs indicated for autoimmune disorders are not necessarily specific to the condition of interest, but rather target the common biological pathways across these diseases. Finally, we provide high scoring drug repurposing candidates that can target common mechanisms involved in type 2 diabetes and Alzheimer’s disease, two conditions that have recently gained attention due to the increased comorbidity among patients.

On the mechanisms of protein interactions: predicting their affinity from unbound tertiary structures.

Motivation: The characterization of the protein-protein association mechanisms is crucial to understanding how biological processes occur. It has been previously shown that the early formation of non-specific encounters enhances the realization of the stereospecific (i.e. native) complex by reducing the dimensionality of the search process. The association rate for the formation of such complex plays a crucial role in the cell biology and depends on how the partners diffuse to be close to each other. Predicting the binding free energy of proteins provides new opportunities to modulate and control protein-protein interactions. However, existing methods require the 3D structure of the complex to predict its affinity, severely limiting their application to interactions with known structures. Results: We present a new approach that relies on the unbound protein structures and protein docking to predict protein-protein binding affinities. Through the study of the docking space (i.e. decoys), the method predicts the binding affinity of the query proteins when the actual structure of the complex itself is unknown. We tested our approach on a set of globular and soluble proteins of the newest affinity benchmark, obtaining accuracy values comparable to other state-of-art methods: a 0.4 correlation coefficient between the experimental and predicted values of ΔG and an error < 3 Kcal/mol. Availability and implementation: The binding affinity predictor is implemented and available at http://sbi.upf.edu/BADock and https://github.com/badocksbi/BADock. Contact: j.planas-iglesias@warwick.ac.uk or baldo.oliva@upf.edu. Supplementary information: Supplementary data are available at Bioinformatics online.