A variational autoencoder (VAE) is a machine learning algorithm, useful for generating a compressed and interpretable latent space. These representations have been generated from various biomedical data types and can be used to produce realistic-looking simulated data. However, standard vanilla VAEs suffer from entangled and uninformative latent spaces, which can be mitigated using other types of VAEs such as β-VAE and MMD-VAE. In this project, we evaluated the ability of VAEs to learn cell morphology characteristics derived from cell images. We trained and evaluated these three VAE variants-Vanilla VAE, β-VAE, and MMD-VAE-on cell morphology readouts and explored the generative capacity of each model to predict compound polypharmacology (the interactions of a drug with more than one target) using an approach called latent space arithmetic (LSA). To test the generalizability of the strategy, we also trained these VAEs using gene expression data of the same compound perturbations and found that gene expression provides complementary information. We found that the β-VAE and MMD-VAE disentangle morphology signals and reveal a more interpretable latent space. We reliably simulated morphology and gene expression readouts from certain compounds thereby predicting cell states perturbed with compounds of known polypharmacology. Inferring cell state for specific drug mechanisms could aid researchers in developing and identifying targeted therapeutics and categorizing off-target effects in the future.

Cytochrome P450 enzymes are found in all life forms. P450s play an important role in drug metabolism, and have potential uses as biocatalysts. Human P450s are membrane-bound proteins. However, the interactions between P450s and their membrane environment are not well-understood. To date, all P450 crystal structures have been obtained from engineered proteins, from which the transmembrane helix was absent. A significant number of computational studies have been performed on P450s, but the majority of these have been performed on the solubilised forms of P450s. Here we present a multiscale approach for modelling P450s, spanning from coarse-grained and atomistic molecular dynamics simulations to reaction modelling using hybrid quantum mechanics/molecular mechanics (QM/MM) methods. To our knowledge, this is the first application of such an integrated multiscale approach to modelling of a membrane-bound enzyme. We have applied this protocol to a key human P450 involved in drug metabolism: CYP3A4. A biologically realistic model of CYP3A4, complete with its transmembrane helix and a membrane, has been constructed and characterised. The dynamics of this complex have been studied, and the oxidation of the anticoagulant R-warfarin has been modelled in the active site. Calculations have also been performed on the soluble form of the enzyme in aqueous solution. Important differences are observed between the membrane and solution systems, most notably for the gating residues and channels that control access to the active site. The protocol that we describe here is applicable to other membrane-bound enzymes.

Due to relatively high costs and labor required for experimental profiling of the full target space of chemical compounds, various machine learning models have been proposed as cost-effective means to advance this process in terms of predicting the most potent compound-target interactions for subsequent verification. However, most of the model predictions lack direct experimental validation in the laboratory, making their practical benefits for drug discovery or repurposing applications largely unknown. Here, we therefore introduce and carefully test a systematic computational-experimental framework for the prediction and pre-clinical verification of drug-target interactions using a well-established kernel-based regression algorithm as the prediction model. To evaluate its performance, we first predicted unmeasured binding affinities in a large-scale kinase inhibitor profiling study, and then experimentally tested 100 compound-kinase pairs. The relatively high correlation of 0.77 (p < 0.0001) between the predicted and measured bioactivities supports the potential of the model for filling the experimental gaps in existing compound-target interaction maps. Further, we subjected the model to a more challenging task of predicting target interactions for such a new candidate drug compound that lacks prior binding profile information. As a specific case study, we used tivozanib, an investigational VEGF receptor inhibitor with currently unknown off-target profile. Among 7 kinases with high predicted affinity, we experimentally validated 4 new off-targets of tivozanib, namely the Src-family kinases FRK and FYN A, the non-receptor tyrosine kinase ABL1, and the serine/threonine kinase SLK. Our sub-sequent experimental validation protocol effectively avoids any possible information leakage between the training and validation data, and therefore enables rigorous model validation for practical applications. These results demonstrate that the kernel-based modeling approach offers practical benefits for probing novel insights into the mode of action of investigational compounds, and for the identification of new target selectivities for drug repurposing applications.

Identifying genetic factors responsible for serious adverse drug reaction (SADR) is of critical importance to personalized medicine. However, genome-wide association studies are hampered due to the lack of case-control samples, and the selection of candidate genes is limited by the lack of understanding of the underlying mechanisms of SADRs. We hypothesize that drugs causing the same type of SADR might share a common mechanism by targeting unexpectedly the same SADR-mediating protein. Hence we propose an approach of identifying the common SADR-targets through constructing and mining an in silico chemical-protein interactome (CPI), a matrix of binding strengths among 162 drug molecules known to cause at least one type of SADR and 845 proteins. Drugs sharing the same SADR outcome were also found to possess similarities in their CPI profiles towards this 845 protein set. This methodology identified the candidate gene of sulfonamide-induced toxic epidermal necrolysis (TEN): all nine sulfonamides that cause TEN were found to bind strongly to MHC I (Cw*4), whereas none of the 17 control drugs that do not cause TEN were found to bind to it. Through an insight into the CPI, we found the Y116S substitution of MHC I (B*5703) enhances the unexpected binding of abacavir to its antigen presentation groove, which explains why B*5701, not B*5703, is the risk allele of abacavir-induced hypersensitivity. In conclusion, SADR targets and the patient-specific off-targets could be identified through a systematic investigation of the CPI, generating important hypotheses for prospective experimental validation of the candidate genes.

Current work in elucidating relationships between diseases has largely been based on pre-existing knowledge of disease genes. Consequently, these studies are limited in their discovery of new and unknown disease relationships. We present the first quantitative framework to compare and contrast diseases by an integrated analysis of disease-related mRNA expression data and the human protein interaction network. We identified 4,620 functional modules in the human protein network and provided a quantitative metric to record their responses in 54 diseases leading to 138 significant similarities between diseases. Fourteen of the significant disease correlations also shared common drugs, supporting the hypothesis that similar diseases can be treated by the same drugs, allowing us to make predictions for new uses of existing drugs. Finally, we also identified 59 modules that were dysregulated in at least half of the diseases, representing a common disease-state "signature". These modules were significantly enriched for genes that are known to be drug targets. Interestingly, drugs known to target these genes/proteins are already known to treat significantly more diseases than drugs targeting other genes/proteins, highlighting the importance of these core modules as prime therapeutic opportunities.

The powerful combination of large-scale drug-related interaction networks and deep learning provides new opportunities for accelerating the process of drug discovery. However, chemical structures that play an important role in drug properties and high-order relations that involve a greater number of nodes are not tackled in current biomedical networks. In this study, we present a general hypergraph learning framework, which introduces Drug-Substructures relationship into Molecular interaction Networks to construct the micro-to-macro drug centric heterogeneous network (DSMN), and develop a multi-branches HyperGraph learning model, called HGDrug, for Drug multi-task predictions. HGDrug achieves highly accurate and robust predictions on 4 benchmark tasks (drug-drug, drug-target, drug-disease, and drug-side-effect interactions), outperforming 8 state-of-the-art task specific models and 6 general-purpose conventional models. Experiments analysis verifies the effectiveness and rationality of the HGDrug model architecture as well as the multi-branches setup, and demonstrates that HGDrug is able to capture the relations between drugs associated with the same functional groups. In addition, our proposed drug-substructure interaction networks can help improve the performance of existing network models for drug-related prediction tasks.

Viruses require host cellular factors for successful replication. A comprehensive systems-level investigation of the virus-host interactome is critical for understanding the roles of host factors with the end goal of discovering new druggable antiviral targets. Gene-trap insertional mutagenesis is a high-throughput forward genetics approach to randomly disrupt (trap) host genes and discover host genes that are essential for viral replication, but not for host cell survival. In this study, we used libraries of randomly mutagenized cells to discover cellular genes that are essential for the replication of 10 distinct cytotoxic mammalian viruses, 1 gram-negative bacterium, and 5 toxins. We herein reported 712 candidate cellular genes, characterizing distinct topological network and evolutionary signatures, and occupying central hubs in the human interactome. Cell cycle phase-specific network analysis showed that host cell cycle programs played critical roles during viral replication (e.g. MYC and TAF4 regulating G0/1 phase). Moreover, the viral perturbation of host cellular networks reflected disease etiology in that host genes (e.g. CTCF, RHOA, and CDKN1B) identified were frequently essential and significantly associated with Mendelian and orphan diseases, or somatic mutations in cancer. Computational drug repositioning framework via incorporating drug-gene signatures from the Connectivity Map into the virus-host interactome identified 110 putative druggable antiviral targets and prioritized several existing drugs (e.g. ajmaline) that may be potential for antiviral indication (e.g. anti-Ebola). In summary, this work provides a powerful methodology with a tight integration of gene-trap insertional mutagenesis testing and systems biology to identify new antiviral targets and drugs for the development of broadly acting and targeted clinical antiviral therapeutics.

One of the main obstacles to the successful treatment of cancer is the phenomenon of drug resistance. A common strategy to overcome resistance is the use of combination therapies. However, the space of possibilities is huge and efficient search strategies are required. Machine Learning (ML) can be a useful tool for the discovery of novel, clinically relevant anti-cancer drug combinations. In particular, deep learning (DL) has become a popular choice for modeling drug combination effects. Here, we set out to examine the impact of different methodological choices on the performance of multimodal DL-based drug synergy prediction methods, including the use of different input data types, preprocessing steps and model architectures. Focusing on the NCI ALMANAC dataset, we found that feature selection based on prior biological knowledge has a positive impact-limiting gene expression data to cancer or drug response-specific genes improved performance. Drug features appeared to be more predictive of drug response, with a 41% increase in coefficient of determination (R2) and 26% increase in Spearman correlation relative to a baseline model that used only cell line and drug identifiers. Molecular fingerprint-based drug representations performed slightly better than learned representations-ECFP4 fingerprints increased R2 by 5.3% and Spearman correlation by 2.8% w.r.t the best learned representations. In general, fully connected feature-encoding subnetworks outperformed other architectures. DL outperformed other ML methods by more than 35% (R2) and 14% (Spearman). Additionally, an ensemble combining the top DL and ML models improved performance by about 6.5% (R2) and 4% (Spearman). Using a state-of-the-art interpretability method, we showed that DL models can learn to associate drug and cell line features with drug response in a biologically meaningful way. The strategies explored in this study will help to improve the development of computational methods for the rational design of effective drug combinations for cancer therapy.

The development of novel therapeutics is urgently required for diseases where existing treatments are failing due to the emergence of resistance. This is particularly pertinent for parasitic infections of the tropics and sub-tropics, referred to collectively as neglected tropical diseases, where the commercial incentives to develop new drugs are weak. One such disease is schistosomiasis, a highly prevalent acute and chronic condition caused by a parasitic helminth infection, with three species of the genus Schistosoma infecting humans. Currently, a single 40-year old drug, praziquantel, is available to treat all infective species, but its use in mass drug administration is leading to signs of drug-resistance emerging. To meet the challenge of developing new therapeutics against this disease, we developed an innovative computational drug repurposing pipeline supported by phenotypic screening. The approach highlighted several protein kinases as interesting new biological targets for schistosomiasis as they play an essential role in many parasite's biological processes. Focusing on this target class, we also report the first elucidation of the kinome of Schistosoma japonicum, as well as updated kinomes of S. mansoni and S. haematobium. In comparison with the human kinome, we explored these kinomes to identify potential targets of existing inhibitors which are unique to Schistosoma species, allowing us to identify novel targets and suggest approved drugs that might inhibit them. These include previously suggested schistosomicidal agents such as bosutinib, dasatinib, and imatinib as well as new inhibitors such as vandetanib, saracatinib, tideglusib, alvocidib, dinaciclib, and 22 newly identified targets such as CHK1, CDC2, WEE, PAKA, MEK1. Additionally, the primary and secondary targets in Schistosoma of those approved drugs are also suggested, allowing for the development of novel therapeutics against this important yet neglected disease.

Although drug resistance in Plasmodium falciparum typically evolves in regions of low transmission, resistance spreads readily following introduction to regions with a heavier disease burden. This suggests that the origin and the spread of resistance are governed by different processes, and that high transmission intensity specifically impedes the origin. Factors associated with high transmission, such as highly immune hosts and competition within genetically diverse infections, are associated with suppression of resistant lineages within hosts. However, interactions between these factors have rarely been investigated and the specific relationship between adaptive immunity and selection for resistance has not been explored. Here, we developed a multiscale, agent-based model of Plasmodium parasites, hosts, and vectors to examine how host and parasite dynamics shape the evolution of resistance in populations with different transmission intensities. We found that selection for antigenic novelty ("immune selection") suppressed the evolution of resistance in high transmission settings. We show that high levels of population immunity increased the strength of immune selection relative to selection for resistance. As a result, immune selection delayed the evolution of resistance in high transmission populations by allowing novel, sensitive lineages to remain in circulation at the expense of the spread of a resistant lineage. In contrast, in low transmission settings, we observed that resistant strains were able to sweep to high population prevalence without interference. Additionally, we found that the relationship between immune selection and resistance changed when resistance was widespread. Once resistance was common enough to be found on many antigenic backgrounds, immune selection stably maintained resistant parasites in the population by allowing them to proliferate, even in untreated hosts, when resistance was linked to a novel epitope. Our results suggest that immune selection plays a role in the global pattern of resistance evolution.

Drug combinations have demonstrated great potential in cancer treatments. They alleviate drug resistance and improve therapeutic efficacy. The fast-growing number of anti-cancer drugs has caused the experimental investigation of all drug combinations to become costly and time-consuming. Computational techniques can improve the efficiency of drug combination screening. Despite recent advances in applying machine learning to synergistic drug combination prediction, several challenges remain. First, the performance of existing methods is suboptimal. There is still much space for improvement. Second, biological knowledge has not been fully incorporated into the model. Finally, many models are lack interpretability, limiting their clinical applications. To address these challenges, we have developed a knowledge-enabled and self-attention transformer boosted deep learning model, TranSynergy, which improves the performance and interpretability of synergistic drug combination prediction. TranSynergy is designed so that the cellular effect of drug actions can be explicitly modeled through cell-line gene dependency, gene-gene interaction, and genome-wide drug-target interaction. A novel Shapley Additive Gene Set Enrichment Analysis (SA-GSEA) method has been developed to deconvolute genes that contribute to the synergistic drug combination and improve model interpretability. Extensive benchmark studies demonstrate that TranSynergy outperforms the state-of-the-art method, suggesting the potential of mechanism-driven machine learning. Novel pathways that are associated with the synergistic combinations are revealed and supported by experimental evidences. They may provide new insights into identifying biomarkers for precision medicine and discovering new anti-cancer therapies. Several new synergistic drug combinations have been predicted with high confidence for ovarian cancer which has few treatment options. The code is available at https://github.com/qiaoliuhub/drug_combination.

Dysregulation of gene expression in Alzheimer's disease (AD) remains elusive, especially at the cell type level. Gene regulatory network, a key molecular mechanism linking transcription factors (TFs) and regulatory elements to govern gene expression, can change across cell types in the human brain and thus serve as a model for studying gene dysregulation in AD. However, AD-induced regulatory changes across brain cell types remains uncharted. To address this, we integrated single-cell multi-omics datasets to predict the gene regulatory networks of four major cell types, excitatory and inhibitory neurons, microglia and oligodendrocytes, in control and AD brains. Importantly, we analyzed and compared the structural and topological features of networks across cell types and examined changes in AD. Our analysis shows that hub TFs are largely common across cell types and AD-related changes are relatively more prominent in some cell types (e.g., microglia). The regulatory logics of enriched network motifs (e.g., feed-forward loops) further uncover cell type-specific TF-TF cooperativities in gene regulation. The cell type networks are also highly modular and several network modules with cell-type-specific expression changes in AD pathology are enriched with AD-risk genes. The further disease-module-drug association analysis suggests cell-type candidate drugs and their potential target genes. Finally, our network-based machine learning analysis systematically prioritized cell type risk genes likely involved in AD. Our strategy is validated using an independent dataset which showed that top ranked genes can predict clinical phenotypes (e.g., cognitive impairment) of AD with reasonable accuracy. Overall, this single-cell network biology analysis provides a comprehensive map linking genes, regulatory networks, cell types and drug targets and reveals cell-type gene dysregulation in AD.

The synergy between human immunodeficiency virus (HIV) and Mycobacterium tuberculosis (MTB) could accelerate the deterioration of immunological functions. Previous studies have explored the pathogenic mechanisms of HIV mono-infection (HMI), MTB mono-infection (MMI) and MTB/HIV co-infection (MHCI), but their similarities and specificities remain to be profoundly investigated. We thus designed a computational framework named IDEN to identify gene pairs related to these states, which were then compared from different perspectives. MMI-related genes showed the highest enrichment level on a greater number of chromosomes. Genes shared by more states tended to be more evolutionarily conserved, posttranslationally modified and topologically important. At the expression level, HMI-specific gene pairs yielded higher correlations, while the overlapping pairs involved in MHCI had significantly lower correlations. The correlation changes of common gene pairs showed that MHCI shared more similarities with MMI. Moreover, MMI- and MHCI-related genes were enriched in more identical pathways and biological processes, further illustrating that MTB may play a dominant role in co-infection. Hub genes specific to each state could promote pathogen infections, while those shared by two states could enhance immune responses. Finally, we improved the network proximity measure for drug repurposing by considering the importance of gene pairs, and approximately ten drug candidates were identified for each disease state.

Receptor tyrosine kinase KIT controls many signal transduction pathways and represents a typical allosterically regulated protein. The mutation-induced deregulation of KIT activity impairs cellular physiological functions and causes serious human diseases. The impact of hotspots mutations (D816H/Y/N/V and V560G/D) localized in crucial regulatory segments, the juxtamembrane region (JMR) and the activation (A-) loop, on KIT internal dynamics was systematically studied by molecular dynamics simulations. The mutational outcomes predicted in silico were correlated with in vitro and in vivo activation rates and drug sensitivities of KIT mutants. The allosteric regulation of KIT in the native and mutated forms is described in terms of communication between the two remote segments, JMR and A-loop. A strong correlation between the communication profile and the structural and dynamical features of KIT in the native and mutated forms was established. Our results provide new insight on the determinants of receptor KIT constitutive activation by mutations and resistance of KIT mutants to inhibitors. Depiction of an intra-molecular component of the communication network constitutes a first step towards an integrated description of vast communication pathways established by KIT in physiopathological contexts.

αβ-tubulin dimers need to convert between a 'bent' conformation observed for free dimers in solution and a 'straight' conformation required for incorporation into the microtubule lattice. Here, we investigate the free energy landscape of αβ-tubulin using molecular dynamics simulations, emphasizing implications for models of assembly, and modulation of the conformational landscape by colchicine, a tubulin-binding drug that inhibits microtubule polymerization. Specifically, we performed molecular dynamics, potential-of-mean force simulations to obtain the free energy profile for unpolymerized GDP-bound tubulin as a function of the ∼12° intradimer rotation differentiating the straight and bent conformers. Our results predict that the unassembled GDP-tubulin heterodimer exists in a continuum of conformations ranging between straight and bent, but, in agreement with existing structural data, suggests that an intermediate bent state has a lower free energy (by ∼1 kcal/mol) and thus dominates in solution. In agreement with predictions of the lattice model of microtubule assembly, lateral binding of two αβ-tubulins strongly shifts the conformational equilibrium towards the straight state, which is then ∼1 kcal/mol lower in free energy than the bent state. Finally, calculations of colchicine binding to a single αβ-tubulin dimer strongly shifts the equilibrium toward the bent states, and disfavors the straight state to the extent that it is no longer thermodynamically populated.

The COVID-19 pandemic has accelerated the need to identify new antiviral therapeutics at pace, including through drug repurposing. We employed a Quadratic Unbounded Binary Optimization (QUBO) model, to search for compounds similar to Remdesivir, the first antiviral against SARS-CoV-2 approved for human use, using a quantum-inspired device. We modelled Remdesivir and compounds present in the DrugBank database as graphs, established the optimal parameters in our algorithm and resolved the Maximum Weighted Independent Set problem within the conflict graph generated. We also employed a traditional Tanimoto fingerprint model. The two methods yielded different lists of lead compounds, with some overlap. While GS-6620 was the top compound predicted by both models, the QUBO model predicted BMS-986094 as second best. The Tanimoto model predicted different forms of cobalamin, also known as vitamin B12. We then determined the half maximal inhibitory concentration (IC50) values in cell culture models of SARS-CoV-2 infection and assessed cytotoxicity. We also demonstrated efficacy against several variants including SARS-CoV-2 Strain England 2 (England 02/2020/407073), B.1.1.7 (Alpha), B.1.351 (Beta) and B.1.617.2 (Delta). Lastly, we employed an in vitro polymerization assay to demonstrate that these compounds directly inhibit the RNA-dependent RNA polymerase (RdRP) of SARS-CoV-2. Together, our data reveal that our QUBO model performs accurate comparisons (BMS-986094) that differed from those predicted by Tanimoto (different forms of vitamin B12); all compounds inhibited replication of SARS-CoV-2 via direct action on RdRP, with both models being useful. While Tanimoto may be employed when performing relatively small comparisons, QUBO is also accurate and may be well suited for very complex problems where computational resources may limit the number and/or complexity of possible combinations to evaluate. Our quantum-inspired screening method can therefore be employed in future searches for novel pharmacologic inhibitors, thus providing an approach for accelerating drug deployment.

Adverse event pathogenesis is often a complex process which compromises multiple events ranging from the molecular to the phenotypic level. In toxicology, Adverse Outcome Pathways (AOPs) aim to formalize this as temporal sequences of events, in which event relationships should be supported by causal evidence according to the tailored Bradford-Hill criteria. One of the criteria is whether events are consistently observed in a certain temporal order and, in this work, we study this time concordance using the concept of "first activation" as data-driven means to generate hypotheses on potentially causal mechanisms. As a case study, we analysed liver data from repeat-dose studies in rats from the TG-GATEs database which comprises measurements across eight timepoints, ranging from 3 hours to 4 weeks post-treatment. We identified time-concordant gene expression-derived events preceding adverse histopathology, which serves as surrogate readout for Drug-Induced Liver Injury (DILI). We find known mechanisms in DILI to be time-concordant, and show further that significance, frequency and log fold change (logFC) of differential expression are metrics which can additionally prioritize events although not necessary to be mechanistically relevant. Moreover, we used the temporal order of transcription factor (TF) expression and regulon activity to identify transcriptionally regulated TFs and subsequently combined this with prior knowledge on functional interactions to derive detailed gene-regulatory mechanisms, such as reduced Hnf4a activity leading to decreased expression and activity of Cebpa. At the same time, also potentially novel events are identified such as Sox13 which is highly significantly time-concordant and shows sustained activation over time. Overall, we demonstrate how time-resolved transcriptomics can derive and support mechanistic hypotheses by quantifying time concordance and how this can be combined with prior causal knowledge, with the aim of both understanding mechanisms of toxicity, as well as potential applications to the AOP framework. We make our results available in the form of a Shiny app (https://anikaliu.shinyapps.io/dili_cascades), which allows users to query events of interest in more detail.

Target-based screening is one of the major approaches in drug discovery. Besides the intended target, unexpected drug off-target interactions often occur, and many of them have not been recognized and characterized. The off-target interactions can be responsible for either therapeutic or side effects. Thus, identifying the genome-wide off-targets of lead compounds or existing drugs will be critical for designing effective and safe drugs, and providing new opportunities for drug repurposing. Although many computational methods have been developed to predict drug-target interactions, they are either less accurate than the one that we are proposing here or computationally too intensive, thereby limiting their capability for large-scale off-target identification. In addition, the performances of most machine learning based algorithms have been mainly evaluated to predict off-target interactions in the same gene family for hundreds of chemicals. It is not clear how these algorithms perform in terms of detecting off-targets across gene families on a proteome scale. Here, we are presenting a fast and accurate off-target prediction method, REMAP, which is based on a dual regularized one-class collaborative filtering algorithm, to explore continuous chemical space, protein space, and their interactome on a large scale. When tested in a reliable, extensive, and cross-gene family benchmark, REMAP outperforms the state-of-the-art methods. Furthermore, REMAP is highly scalable. It can screen a dataset of 200 thousands chemicals against 20 thousands proteins within 2 hours. Using the reconstructed genome-wide target profile as the fingerprint of a chemical compound, we predicted that seven FDA-approved drugs can be repurposed as novel anti-cancer therapies. The anti-cancer activity of six of them is supported by experimental evidences. Thus, REMAP is a valuable addition to the existing in silico toolbox for drug target identification, drug repurposing, phenotypic screening, and side effect prediction. The software and benchmark are available at https://github.com/hansaimlim/REMAP.

Systematic identification of protein-drug interaction networks is crucial to correlate complex modes of drug action to clinical indications. We introduce a novel computational strategy to identify protein-ligand binding profiles on a genome-wide scale and apply it to elucidating the molecular mechanisms associated with the adverse drug effects of Cholesteryl Ester Transfer Protein (CETP) inhibitors. CETP inhibitors are a new class of preventive therapies for the treatment of cardiovascular disease. However, clinical studies indicated that one CETP inhibitor, Torcetrapib, has deadly off-target effects as a result of hypertension, and hence it has been withdrawn from phase III clinical trials. We have identified a panel of off-targets for Torcetrapib and other CETP inhibitors from the human structural genome and map those targets to biological pathways via the literature. The predicted protein-ligand network is consistent with experimental results from multiple sources and reveals that the side-effect of CETP inhibitors is modulated through the combinatorial control of multiple interconnected pathways. Given that combinatorial control is a common phenomenon observed in many biological processes, our findings suggest that adverse drug effects might be minimized by fine-tuning multiple off-target interactions using single or multiple therapies. This work extends the scope of chemogenomics approaches and exemplifies the role that systems biology has in the future of drug discovery.

Progress in systems medicine brings promise to addressing patient heterogeneity and individualized therapies. Recently, genome-scale models of metabolism have been shown to provide insight into the mechanistic link between drug therapies and systems-level off-target effects while being expanded to explicitly include the three-dimensional structure of proteins. The integration of these molecular-level details, such as the physical, structural, and dynamical properties of proteins, notably expands the computational description of biochemical network-level properties and the possibility of understanding and predicting whole cell phenotypes. In this study, we present a multi-scale modeling framework that describes biological processes which range in scale from atomistic details to an entire metabolic network. Using this approach, we can understand how genetic variation, which impacts the structure and reactivity of a protein, influences both native and drug-induced metabolic states. As a proof-of-concept, we study three enzymes (catechol-O-methyltransferase, glucose-6-phosphate dehydrogenase, and glyceraldehyde-3-phosphate dehydrogenase) and their respective genetic variants which have clinically relevant associations. Using all-atom molecular dynamic simulations enables the sampling of long timescale conformational dynamics of the proteins (and their mutant variants) in complex with their respective native metabolites or drug molecules. We find that changes in a protein's structure due to a mutation influences protein binding affinity to metabolites and/or drug molecules, and inflicts large-scale changes in metabolism.