Inferring gene targets of drugs and chemical compounds from gene expression profiles

High-throughput platform for yeast morphological profiling predicts the targets of bioactive compounds

Morphological profiling is an omics-based approach for predicting intracellular targets of chemical compounds in which the dose-dependent morphological changes induced by the compound are systematically compared to the morphological changes in gene-deleted cells. In this study, we developed a reliable high-throughput (HT) platform for yeast morphological profiling using drug-hypersensitive strains to minimize compound use, HT microscopy to speed up data generation and analysis, and a generalized linear model to predict targets with high reliability. We first conducted a proof-of-concept study using six compounds with known targets: bortezomib, hydroxyurea, methyl methanesulfonate, benomyl, tunicamycin, and echinocandin B. Then we applied our platform to predict the mechanism of action of a novel diferulate-derived compound, poacidiene. Morphological profiling of poacidiene implied that it affects the DNA damage response, which genetic analysis confirmed. Furthermore, we found that poacidiene inhibits the growth of phytopathogenic fungi, implying applications as an effective antifungal agent. Thus, our platform is a new whole-cell target prediction tool for drug discovery.

Drug target inference by mining transcriptional data using a novel graph convolutional network framework

A fundamental challenge that arises in biomedicine is the need to characterize compounds in a relevant cellular context in order to reveal potential on-target or off-target effects. Recently, the fast accumulation of gene transcriptional profiling data provides us an unprecedented opportunity to explore the protein targets of chemical compounds from the perspective of cell transcriptomics and RNA biology. Here, we propose a novel Siamese spectral-based graph convolutional network (SSGCN) model for inferring the protein targets of chemical compounds from gene transcriptional profiles. Although the gene signature of a compound perturbation only provides indirect clues of the interacting targets, and the biological networks under different experiment conditions further complicate the situation, the SSGCN model was successfully trained to learn from known compound-target pairs by uncovering the hidden correlations between compound perturbation profiles and gene knockdown profiles. On a benchmark set and a large time-split validation dataset, the model achieved higher target inference accuracy as compared to previous methods such as Connectivity Map. Further experimental validations of prediction results highlight the practical usefulness of SSGCN in either inferring the interacting targets of compound, or reversely, in finding novel inhibitors of a given target of interest.

DeltaNeTS+: elucidating the mechanism of drugs and diseases using gene expression and transcriptional regulatory networks

BACKGROUND

Knowledge on the molecular targets of diseases and drugs is crucial for elucidating disease pathogenesis and mechanism of action of drugs, and for driving drug discovery and treatment formulation. In this regard, high-throughput gene transcriptional profiling has become a leading technology, generating whole-genome data on the transcriptional alterations caused by diseases or drug compounds. However, identifying direct gene targets, especially in the background of indirect (downstream) effects, based on differential gene expressions is difficult due to the complexity of gene regulatory network governing the gene transcriptional processes.

RESULTS

In this work, we developed a network analysis method, called DeltaNeTS+, for inferring direct gene targets of drugs and diseases from gene transcriptional profiles. DeltaNeTS+ uses a gene regulatory network model to identify direct perturbations to the transcription of genes using gene expression data. Importantly, DeltaNeTS+ is able to combine both steady-state and time-course expression profiles, as well as leverage information on the gene network structure. We demonstrated the power of DeltaNeTS+ in predicting gene targets using gene expression data in complex organisms, including Caenorhabditis elegans and human cell lines (T-cell and Calu-3). More specifically, in an application to time-course gene expression profiles of influenza A H1N1 (swine flu) and H5N1 (avian flu) infection, DeltaNeTS+ shed light on the key differences of dynamic cellular perturbations caused by the two influenza strains.

CONCLUSION

DeltaNeTS+ is a powerful network analysis tool for inferring gene targets from gene expression profiles. As demonstrated in the case studies, by incorporating available information on gene network structure, DeltaNeTS+ produces accurate predictions of direct gene targets from a small sample size (~ 10 s). Integrating static and dynamic expression data with transcriptional network structure extracted from genomic information, as enabled by DeltaNeTS+, is crucial toward personalized medicine, where treatments can be tailored to individual patients. DeltaNeTS+ can be freely downloaded from http://www.github.com/cabsel/deltanetsplus .

Applications of Machine Learning in Drug Target Discovery

Drug target discovery is a critical step in drug development. It is the basis of modern drug development because it determines the target molecules related to specific diseases in advance. Predicting drug targets by computational methods saves a great deal of financial and material resources compared to in vitro experiments. Therefore, several computational methods for drug target discovery have been designed. Recently, machine learning (ML) methods in biomedicine have developed rapidly. In this paper, we present an overview of drug target discovery methods based on machine learning. Considering that some machine learning methods integrate network analysis to predict drug targets, network-based methods are also introduced in this article. Finally, the challenges and future outlook of drug target discovery are discussed.

TREAP: A New Topological Approach to Drug Target Inference

Over 50% of drugs fail in stage 3 clinical trials, many because of a poor understanding of the drug's mechanisms of action (MoA). A better comprehension of drug MoA will significantly improve research and development (R&D). Current proposed algorithms, such as ProTINA and DeMAND, can be overly complex. Additionally, they are unable to predict whether the drug-induced gene expression or the topology of the networks used to model gene regulation primarily impacts accurate drug target inference. In this work, we evaluate how network and gene expression data affect ProTINA's accuracy. We find that network topology predominantly determines the accuracy of ProTINA's predictions. We further show that the size of an interaction network and/or selecting cell-specific networks has a limited effect on accuracy. We then demonstrate that a specific network topology measure, betweenness, can be used to improve drug target prediction. Based on these results, we create a new algorithm, TREAP, that combines betweenness values and adjusted p-values for target inference. TREAP offers an alternative approach to drug target inference and is advantageous because it is not computationally demanding, provides easy-to-interpret results, and is often more accurate at predicting drug targets than current state-of-the-art approaches.

Identifying "Many-to-Many" Relationships between Gene-Expression Data and Drug-Response Data via Sparse Binary Matching

Identifying gene-drug patterns is a critical step in pharmacology for unveiling disease mechanisms and drug discovery. The availability of high-throughput technologies accumulates massive large-scale pharmacological and genomic data, and thus provides a new substantial opportunity to deeply understand how the oncogenic genes and the therapeutic drugs relate to each other. However, most previous studies merely used the pharmacological and genomic datasets without any prior knowledge to infer the gene-drug patterns. Here, we proposed a novel network-guided sparse binary matching model (NSBM) to decode these relationships hidden in the datasets. Not only the large-scale gene-expression data and drug-response data are jointly analyzed in our method, but also the additional prior information of genes and drugs are integrated into the form of network-based regularization. The essential structure of the NSBM model is a convex quadratic minimization problem with network-based penalties. It was demonstrated to be superior when compared with two benchmark methods through extensive experiments on both synthetic and empirical data. Posterior validation, including gene-ontology and enrichment analysis, confirmed the effectiveness of NSBM in revealing gene-drug patterns on a large-scale heterogeneous data source.

Network perturbation analysis of gene transcriptional profiles reveals protein targets and mechanism of action of drugs and influenza A viral infection

Genome-wide transcriptional profiling provides a global view of cellular state and how this state changes under different treatments (e.g. drugs) or conditions (e.g. healthy and diseased). Here, we present ProTINA (Protein Target Inference by Network Analysis), a network perturbation analysis method for inferring protein targets of compounds from gene transcriptional profiles. ProTINA uses a dynamic model of the cell-type specific protein-gene transcriptional regulation to infer network perturbations from steady state and time-series differential gene expression profiles. A candidate protein target is scored based on the gene network's dysregulation, including enhancement and attenuation of transcriptional regulatory activity of the protein on its downstream genes, caused by drug treatments. For benchmark datasets from three drug treatment studies, ProTINA was able to provide highly accurate protein target predictions and to reveal the mechanism of action of compounds with high sensitivity and specificity. Further, an application of ProTINA to gene expression profiles of influenza A viral infection led to new insights of the early events in the infection.

A dual controllability analysis of influenza virus-host protein-protein interaction networks for antiviral drug target discovery

Background

Host factors of influenza virus replication are often found in key topological positions within protein-protein interaction networks. This work explores how protein states can be manipulated through controllability analysis: the determination of the minimum manipulation needed to drive the cell system to any desired state. Here, we complete a two-part controllability analysis of two protein networks: a host network representing the healthy cell state and an influenza A virus-host network representing the infected cell state. In this context, controllability analyses aim to identify key regulating host factors of the infected cell’s progression. This knowledge can be utilized in further biological analysis to understand disease dynamics and isolate proteins for study as drug target candidates.

Results

Both topological and controllability analyses provide evidence of wide-reaching network effects stemming from the addition of viral-host protein interactions. Virus interacting and driver host proteins are significant both topologically and in controllability, therefore playing important roles in cell behavior during infection. Functional analysis finds overlap of results with previous siRNA studies of host factors involved in influenza replication, NF-kB pathway and infection relevance, and roles as interferon regulating genes. 24 proteins are identified as holding regulatory roles specific to the infected cell by measures of topology, controllability, and functional role. These proteins are recommended for further study as potential antiviral drug targets.

Conclusions

Seasonal outbreaks of influenza A virus are a major cause of illness and death around the world each year with a constant threat of pandemic infection. This research aims to increase the efficiency of antiviral drug target discovery using existing protein-protein interaction data and network analysis methods. These results are beneficial to future studies of influenza virus, both experimental and computational, and provide evidence that the combination of topology and controllability analyses may be valuable for future efforts in drug target discovery.

Electronic supplementary material

The online version of this article (10.1186/s12859-019-2917-z) contains supplementary material, which is available to authorized users.

Drug candidate identification based on gene expression of treated cells using tensor decomposition-based unsupervised feature extraction for large-scale data

Background

Although in silico drug discovery is necessary for drug development, two major strategies, a structure-based and ligand-based approach, have not been completely successful. Currently, the third approach, inference of drug candidates from gene expression profiles obtained from the cells treated with the compounds under study requires the use of a training dataset. Here, the purpose was to develop a new approach that does not require any pre-existing knowledge about the drug–protein interactions, but these interactions can be inferred by means of an integrated approach using gene expression profiles obtained from the cells treated with the analysed compounds and the existing data describing gene–gene interactions.

Results

In the present study, using tensor decomposition-based unsupervised feature extraction, which represents an extension of the recently proposed principal-component analysis-based feature extraction, gene sets and compounds with a significant dose-dependent activity were screened without any training datasets. Next, after these results were combined with the data showing perturbations in single-gene expression profiles, genes targeted by the analysed compounds were inferred. The set of target genes thus identified was shown to significantly overlap with known target genes of the compounds under study.

Conclusions

The method is specifically designed for large-scale datasets (including hundreds of treatments with compounds), not for conventional small-scale datasets. The obtained results indicate that two compounds that have not been extensively studied, WZ-3105 and CGP-60474, represent promising drug candidates targeting multiple cancers, including melanoma, adenocarcinoma, liver carcinoma, and breast, colon, and prostate cancers, which were analysed in this in silico study.

Electronic supplementary material

The online version of this article (10.1186/s12859-018-2395-8) contains supplementary material, which is available to authorized users.

In Silico Computational Transcriptomics Reveals Novel Endocrine Disruptors in Largemouth Bass ( Micropterus salmoides)

In recent years, decreases in fish populations have been attributed, in part, to the effect of environmental chemicals on ovarian development. To understand the underlying molecular events we developed a dynamic model of ovary development linking gene transcription to key physiological end points, such as gonadosomatic index (GSI), plasma levels of estradiol (E2) and vitellogenin (VTG), in largemouth bass ( Micropterus salmoides). We were able to identify specific clusters of genes, which are affected at different stages of ovarian development. A subnetwork was identified that closely linked gene expression and physiological end points and by interrogating the Comparative Toxicogenomic Database (CTD), quercetin and tretinoin (ATRA) were identified as two potential candidates that may perturb this system. Predictions were validated by investigation of reproductive associated transcripts using qPCR in ovary and in the liver of both male and female largemouth bass treated after a single injection of quercetin and tretinoin (10 and 100 μg/kg). Both compounds were found to significantly alter the expression of some of these genes. Our findings support the use of omics and online repositories for identification of novel, yet untested, compounds. This is the first study of a dynamic model that links gene expression patterns across stages of ovarian development.

BGRMI: A method for inferring gene regulatory networks from time-course gene expression data and its application in breast cancer research

Reconstructing gene regulatory networks (GRNs) from gene expression data is a challenging problem. Existing GRN reconstruction algorithms can be broadly divided into model-free and model–based methods. Typically, model-free methods have high accuracy but are computation intensive whereas model-based methods are fast but less accurate. We propose Bayesian Gene Regulation Model Inference (BGRMI), a model-based method for inferring GRNs from time-course gene expression data. BGRMI uses a Bayesian framework to calculate the probability of different models of GRNs and a heuristic search strategy to scan the model space efficiently. Using benchmark datasets, we show that BGRMI has higher/comparable accuracy at a fraction of the computational cost of competing algorithms. Additionally, it can incorporate prior knowledge of potential gene regulation mechanisms and TF hetero-dimerization processes in the GRN reconstruction process. We incorporated existing ChIP-seq data and known protein interactions between TFs in BGRMI as sources of prior knowledge to reconstruct transcription regulatory networks of proliferating and differentiating breast cancer (BC) cells from time-course gene expression data. The reconstructed networks revealed key driver genes of proliferation and differentiation in BC cells. Some of these genes were not previously studied in the context of BC, but may have clinical relevance in BC treatment.