Chemical-genetic Interactions Research Articles

A phenotypic drug discovery approach by latent interaction in deep learning

Contemporary drug discovery paradigms rely heavily on binding assays about the bio-physicochemical processes. However, this dominant approach suffers from overlooked higher-order interactions arising from the intricacies of molecular mechanisms, such as those involving cis -regulatory elements. It introduces potential impairments and restrains the potential development of computational methods. To address this limitation, I developed a deep learning model that leverages an end-to-end approach, relying exclusively on therapeutic information about drugs. By transforming textual representations of drug and virus genetic information into high-dimensional latent representations, this method evades the challenges arising from insufficient information about binding specificities. Its strengths lie in its ability to implicitly consider complexities such as epistasis and chemical–genetic interactions, and to handle the pervasive challenge of data scarcity. Through various modeling skills and data augmentation techniques, the proposed model demonstrates outstanding performance in out-of-sample validations, even in scenarios with unknown complex interactions. Furthermore, the study highlights the importance of chemical diversity for model training. While the method showcases the feasibility of deep learning in data-scarce scenarios, it reveals a promising alternative for drug discovery in situations where knowledge of underlying mechanisms is limited.

Chemical genetic interactions elucidate pathways controlling tuberculosis antibiotic efficacy during infection.

The inevitable rise of multi-drug-resistant tuberculosis underscores the urgent need for new TB drugs and novel drug targets while prioritizing synergistic drug combinations. Chemical-genetic interaction (CGI) studies have delineated bacterial pathways influencing antibiotic efficacy and uncovered druggable pathways that synergize with TB drugs. However, most studies are conducted in vitro , limiting our understanding of how the host environment influences drug-mutant interactions. Using an inducible mutant library targeting essential Mtb genes to characterize CGIs during infection, this study reveals that CGIs are both drug-specific and drug-agnostic and differ significantly from those observed in vitro . Synergistic CGIs comprised distinct metabolic pathways mediating antibiotic efficacy, revealing novel drug mechanisms of action, and defining potential drug targets that would synergize with frontline antitubercular drugs.

CRISPR Screening and Comparative LC-MS Analysis Identify Genes Mediating Efficacy of Delamanid and Pretomanid against Mycobacterium tuberculosis.

Tuberculosis (TB), the leading cause of death from bacterial infections worldwide, results from infection with Mycobacterium tuberculosis (Mtb). The antitubercular agents delamanid (DLM) and pretomanid (PMD) are nitroimidazole prodrugs that require activation by an enzyme intrinsic to Mtb; however, the mechanism(s) of action and the associated metabolic pathways are largely unclear. Profiling of the chemical-genetic interactions of PMD and DLM in Mtb using combined CRISPR screening reveals that the mutation of rv2073c increases susceptibility of Mtb to these nitroimidazole drugs both in vitro and in infected mice, whereas mutation of rv0078 increases drug resistance. Further assays show that Rv2073c might confer intrinsic resistance to DLM/PMD by interfering with inhibition of the drug target, decaprenylphophoryl-2-keto-b-D-erythro-pentose reductase (DprE2), by active nicotinamide adenine dinucleotide (NAD) adducts. Characterization of the metabolic pathways of DLM/PMD in Mtb using a combination of chemical genetics and comparative liquid chromatography-mass spectrometry (LC-MS) analysis of DLM/PMD metabolites reveals that Rv0077c, which is negatively regulated by Rv0078, mediates drug resistance by metabolizing activated DLM/PMD. These results might guide development of new nitroimidazole prodrugs and new regimens for TB treatment.

CLCC1 promotes hepatic neutral lipid flux and nuclear pore complex assembly.

Imbalances in lipid storage and secretion lead to the accumulation of hepatocyte lipid droplets (LDs) (i.e., hepatic steatosis). Our understanding of the mechanisms that govern the channeling of hepatocyte neutral lipids towards cytosolic LDs or secreted lipoproteins remains incomplete. Here, we performed a series of CRISPR-Cas9 screens under different metabolic states to uncover mechanisms of hepatic neutral lipid flux. Clustering of chemical-genetic interactions identified CLIC-like chloride channel 1 (CLCC1) as a critical regulator of neutral lipid storage and secretion. Loss of CLCC1 resulted in the buildup of large LDs in hepatoma cells and knockout in mice caused liver steatosis. Remarkably, the LDs are in the lumen of the ER and exhibit properties of lipoproteins, indicating a profound shift in neutral lipid flux. Finally, remote homology searches identified a domain in CLCC1 that is homologous to yeast Brl1p and Brr6p, factors that promote the fusion of the inner and outer nuclear envelopes during nuclear pore complex assembly. Loss of CLCC1 lead to extensive nuclear membrane herniations, consistent with impaired nuclear pore complex assembly. Thus, we identify CLCC1 as the human Brl1p/Brr6p homolog and propose that CLCC1-mediated membrane remodeling promotes hepatic neutral lipid flux and nuclear pore complex assembly.

Harnessing genetic interactions for prediction of immune checkpoint inhibitors response signature in cancer cells

Genetic interactions (GIs) refer to two altered genes having a combined effect that is not seen individually. They play a crucial role in influencing drug efficacy. We utilized CGIdb 2.0 (http://www.medsysbio.org/CGIdb2/), an updated database of comprehensively published GIs information, encompassing synthetic lethality (SL), synthetic viability (SV), and chemical-genetic interactions. CGIdb 2.0 elucidates GIs relationships between or within protein complex models by integrating protein-protein physical interactions. Additionally, we introduced GENIUS (GENetic Interactions mediated drUg Signature) to leverage GIs for identifying the response signature of immune checkpoint inhibitors (ICIs). GENIUS identified high MAP4K4 expression as a resistant signature and high HERC4 expression as a sensitive signature for ICIs treatment. Melanoma patients with high expression of MAP4K4 were associated with decreased efficacy and poorer survival following ICIs treatment. Conversely, overexpression of HERC4 in melanoma patients correlated with a positive response to ICIs. Notably, HERC4 enhances sensitivity to immunotherapy by facilitating antigen presentation. Analyses of immune cell infiltration and single-cell data revealed that B cells expressing MAP4K4 may contribute to resistance to ICIs in melanoma. Overall, CGIdb 2.0, provides integrated GIs data, thus serving as a crucial tool for exploring drug effects.

A dose-response model for statistical analysis of chemical genetic interactions in CRISPRi screens.

An important application of CRISPR interference (CRISPRi) technology is for identifying chemical-genetic interactions (CGIs). Discovery of genes that interact with exposure to antibiotics can yield insights to drug targets and mechanisms of action or resistance. The objective is to identify CRISPRi mutants whose relative abundance is suppressed (or enriched) in the presence of a drug when the target protein is depleted, reflecting synergistic behavior. Different sgRNAs for a given target can induce a wide range of protein depletion and differential effects on growth rate. The effect of sgRNA strength can be partially predicted based on sequence features. However, the actual growth phenotype depends on the sensitivity of cells to depletion of the target protein. For essential genes, sgRNA efficiency can be empirically measured by quantifying effects on growth rate. We observe that the most efficient sgRNAs are not always optimal for detecting synergies with drugs. sgRNA efficiency interacts in a non-linear way with drug sensitivity, producing an effect where the concentration-dependence is maximized for sgRNAs of intermediate strength (and less so for sgRNAs that induce too much or too little target depletion). To capture this interaction, we propose a novel statistical method called CRISPRi-DR (for Dose-Response model) that incorporates both sgRNA efficiencies and drug concentrations in a modified dose-response equation. We use CRISPRi-DR to re-analyze data from a recent CGI experiment in Mycobacterium tuberculosis to identify genes that interact with antibiotics. This approach can be generalized to non-CGI datasets, which we show via an CRISPRi dataset for E. coli growth on different carbon sources. The performance is competitive with the best of several related analytical methods. However, for noisier datasets, some of these methods generate far more significant interactions, likely including many false positives, whereas CRISPRi-DR maintains higher precision, which we observed in both empirical and simulated data.

DRSPRING: Graph convolutional network (GCN)-Based drug synergy prediction utilizing drug-induced gene expression profile

Great efforts have been made over the years to identify novel drug pairs with synergistic effects. Although numerous computational approaches have been proposed to analyze diverse types of biological big data, the pharmacogenomic profiles, presumably the most direct proxy of drug effects, have been rarely used due to the data sparsity problem. In this study, we developed a composite deep-learning-based model that predicts the drug synergy effect utilizing pharmacogenomic profiles as well as molecular properties. Graph convolutional network (GCN) was used to represent and integrate the chemical structure, genetic interactions, drug-target information, and gene expression profiles of cell lines. Insufficient amount of pharmacogenomic data, i.e., drug-induced expression profiles from the LINCS project, was resolved by augmenting the data with the predicted profiles. Our method learned and predicted the Loewe synergy score in the DrugComb database and achieved a better or comparable performance compared to other published methods in a benchmark test. We also investigated contribution of various input features, which highlighted the value of basal gene expression and pharmacogenomic profiles of each cell line. Importantly, DRSPRING (DRug Synergy PRediction by INtegrated GCN) can be applied to any drug pairs and any cell lines, greatly expanding its applicability compared to previous methods.

Breaking down barriers: comprehensive functional analysis of the Aspergillus niger chitin synthase repertoire

BackgroundMembers of the fungal kingdom are heterotrophic eukaryotes encased in a chitin containing cell wall. This polymer is vital for cell wall stiffness and, ultimately, cell shape. Most fungal genomes contain numerous putative chitin synthase encoding genes. However, systematic functional analysis of the full chitin synthase catalogue in a given species is rare. This greatly limits fundamental understanding and potential applications of manipulating chitin synthesis across the fungal kingdom.ResultsIn this study, we conducted in silico profiling and subsequently deleted all predicted chitin synthase encoding genes in the multipurpose cell factory Aspergillus niger. Phylogenetic analysis suggested nine chitin synthases evolved as three distinct groups. Transcript profiling and co-expression network construction revealed remarkably independent expression, strongly supporting specific role(s) for the respective chitin synthases. Deletion mutants confirmed all genes were dispensable for germination, yet impacted colony spore titres, chitin content at hyphal septa, and internal architecture of submerged fungal pellets. We were also able to assign specific roles to individual chitin synthases, including those impacting colony radial growth rates (ChsE, ChsF), lateral cell wall chitin content (CsmA), chemical genetic interactions with a secreted antifungal protein (CsmA, CsmB, ChsE, ChsF), resistance to therapeutics (ChsE), and those that modulated pellet diameter in liquid culture (ChsA, ChsB). From an applied perspective, we show chsF deletion increases total protein in culture supernatant over threefold compared to the control strain, indicating engineering filamentous fungal chitin content is a high priority yet underexplored strategy for strain optimization.ConclusionThis study has conducted extensive analysis for the full chitin synthase encoding gene repertoire of A. niger. For the first time we reveal both redundant and non-redundant functional roles of chitin synthases in this fungus. Our data shed light on the complex, multifaceted, and dynamic role of chitin in fungal growth, morphology, survival, and secretion, thus improving fundamental understanding and opening new avenues for biotechnological applications in fungi.

Cytidine deaminases APOBEC3C and APOBEC3D promote DNA replication stress resistance in pancreatic cancer cells.

Gemcitabine is a potent inhibitor of DNA replication and is a mainstay therapeutic for diverse cancers, particularly pancreatic ductal adenocarcinoma (PDAC). However, most tumors remain refractory to gemcitabine therapies. Here, to define the cancer cell response to gemcitabine, we performed genome-scale CRISPR-Cas9 chemical-genetic screens in PDAC cells and found selective loss of cell fitness upon disruption of the cytidine deaminases APOBEC3C and APOBEC3D. Following gemcitabine treatment, APOBEC3C and APOBEC3D promote DNA replication stress resistance and cell survival by deaminating cytidines in the nuclear genome to ensure DNA replication fork restart and repair in PDAC cells. We provide evidence that the chemical-genetic interaction between APOBEC3C or APOBEC3D and gemcitabine is absent in nontransformed cells but is recapitulated across different PDAC cell lines, in PDAC organoids and in PDAC xenografts. Thus, we uncover roles for APOBEC3C and APOBEC3D in DNA replication stress resistance and offer plausible targets for improving gemcitabine-based therapies for PDAC.

A scalable platform for efficient CRISPR-Cas9 chemical-genetic screens of DNA damage-inducing compounds

Current approaches to define chemical-genetic interactions (CGIs) in human cell lines are resource-intensive. We designed a scalable chemical-genetic screening platform by generating a DNA damage response (DDR)-focused custom sgRNA library targeting 1011 genes with 3033 sgRNAs. We performed five proof-of-principle compound screens and found that the compounds’ known modes-of-action (MoA) were enriched among the compounds’ CGIs. These scalable screens recapitulated expected CGIs at a comparable signal-to-noise ratio (SNR) relative to genome-wide screens. Furthermore, time-resolved CGIs, captured by sequencing screens at various time points, suggested an unexpected, late interstrand-crosslinking (ICL) repair pathway response to camptothecin-induced DNA damage. Our approach can facilitate screening compounds at scale with 20-fold fewer resources than commonly used genome-wide libraries and produce biologically informative CGI profiles.

Identification of potential inhibitors of casein kinase 2 alpha of Plasmodium falciparum with potent in vitro activity.

Drug resistance to commercially available antimalarials is a major obstacle in malaria control and elimination, creating the need to find new antiparasitic compounds with novel mechanisms of action. The success of kinase inhibitors for oncological treatments has paved the way for the exploitation of protein kinases as drug targets in various diseases, including malaria. Casein kinases are ubiquitous serine/threonine kinases involved in a wide range of cellular processes such as mitotic checkpoint signaling, DNA damage response, and circadian rhythm. In Plasmodium, it is suggested that these protein kinases are essential for both asexual and sexual blood-stage parasites, reinforcing their potential as targets for multi-stage antimalarials. To identify new putative PfCK2α inhibitors, we utilized an in silico chemogenomic strategy involving virtual screening with docking simulations and quantitative structure-activity relationship predictions. Our investigation resulted in the discovery of a new quinazoline molecule (542), which exhibited potent activity against asexual blood stages and a high selectivity index (>100). Subsequently, we conducted chemical-genetic interaction analysis on yeasts with mutations in casein kinases. Our chemical-genetic interaction results are consistent with the hypothesis that 542 inhibits yeast Cka1, which has a hinge region with high similarity to PfCK2α. This finding is in agreement with our in silico results suggesting that 542 inhibits PfCK2α via hinge region interaction.

CRISPR-Cas9 as a Potential Cancer Therapy Agent: An Update

Cancer is the second leading cause of death globally and remains a major economic and social burden. Although our understanding of cancer at the molecular level continues to improve, more effort is needed to develop new therapeutic tools and approaches exploiting these advances. Due to its high efficiency and accuracy, the CRISPR-Cas9 genome editing technique has recently emerged as a cancer treatment strategy. Among its many applications, CRISPR-Cas9 has shown an unprecedented clinical potential to discover novel targets for cancer therapy and to dissect chemical-genetic interactions, providing insight into how tumors respond to drug treatment. Moreover, CRISPR-Cas9 can be employed to rapidly engineer immune cells and oncolytic viruses for cancer immunotherapeutic applications. More importantly, the ability of CRISPR-Cas9 to accurately edit genes, not only in cell culture models and model organisms but also in humans, allows its use in therapeutic explorations this review, important considerations for the use of CRISPR/Cas9 in therapeutic properties are discussed, along with major challenges that will need to be addressed before clinical examinations for a complex and polygenic disease such as cancer. This review aimed to explore the potential of the CRISPR-Cas9 genome editing technique as a cancer treatment strategy. Specifically, we will discuss how CRISPR-Cas9 can be used to discover novel targets for cancer therapy and to dissect chemical-genetic interactions.

RNAi screening reveals a synthetic chemical-genetic interaction between ATP synthase and PFK1 in cancer cells.

To meet cellular bioenergetic and biosynthetic demands, cancer cells remodel their metabolism to increase glycolytic flux, a phenomenon known as the Warburg effect and believed to contribute to cancer malignancy. Among glycolytic enzymes, phosphofructokinase-1 (PFK1) has been shown to act as a rate-limiting enzyme and to facilitate the Warburg effect in cancer cells. In this study, however, we found that decreased PFK1 activity did not affect cell survival or proliferation in cancer cells. This raised a question regarding the importance of PFK1 in malignancy. To gain insights into the role of PFK1 in cancer metabolism and the possibility of adopting it as a novel anticancer therapeutic target, we screened for genes that caused lethality when they were knocked down in the presence of tryptolinamide (TLAM), a PFK1 inhibitor. The screen revealed a synthetic chemical-genetic interaction between genes encoding subunits of ATP synthase (complex V) and TLAM. Indeed, after TLAM treatment, the sensitivity of HeLa cells to oligomycin A (OMA), an ATP synthase inhibitor, was 13,000 times higher than that of untreated cells. Furthermore, this sensitivity potentiation by TLAM treatment was recapitulated by genetic mutations of PFK1. By contrast, TLAM did not potentiate the sensitivity of normal fibroblast cell lines to OMA, possibly due to their reduced energy demands compared to cancer cells. We also showed that the PFK1-mediated glycolytic pathway can act as an energy reservoir. Selective potentiation of the efficacy of ATP synthase inhibitors by PFK1 inhibition may serve as a foundation for novel anticancer therapeutic strategies.

Application of combined CRISPR screening for genetic and chemical-genetic interaction profiling in Mycobacterium tuberculosis.

CRISPR screening, including CRISPR interference (CRISPRi) and CRISPR-knockout (CRISPR-KO) screening, has become a powerful technology in the genetic screening of eukaryotes. In contrast with eukaryotes, CRISPR-KO screening has not yet been applied to functional genomics studies in bacteria. Here, we constructed genome-scale CRISPR-KO and also CRISPRi libraries in Mycobacterium tuberculosis (Mtb). We first examined these libraries to identify genes essential for Mtb viability. Subsequent screening identified dozens of genes associated with resistance/susceptibility to the antitubercular drug bedaquiline (BDQ). Genetic and chemical validation of the screening results suggested that it provided a valuable resource to investigate mechanisms of action underlying the effects of BDQ and to identify chemical-genetic synergies that can be used to optimize tuberculosis therapy. In summary, our results demonstrate the potential for efficient genome-wide CRISPR-KO screening in bacteria and establish a combined CRISPR screening approach for high-throughput investigation of genetic and chemical-genetic interactions in Mtb.

CSIG-16. EXPLOITING A NEGATIVE FEEDBACK LOOP LINKING S6 KINASES AND AXL TO REDUCE PYRIMIDINE BIOSYNTHESIS IN PTEN-DEFICIENT GLIOBLASTOMA

Abstract PTEN inactivation triggers oncodependent signal transduction through the PI3K/Akt signal transduction pathway in glioblastoma (GBM). Targeting the PI3K/Akt pathway to counteract PTEN loss in GBM has been impeded by negative feedback signal transduction networks mediated by the downstream protein kinases S6K1 and S6K2. Here, we show that the receptor tyrosine kinase AXL is a major target of that S6K feedback signaling and that combined inactivation of just S6K1 and AXL is an effective therapeutic strategy for treatment of PTEN-deficient GBM. Chemical-genetic interaction studies in gliomasphere and GBM cell lines revealed critical and independent roles for S6K1 and S6K2 in mediating GBM growth in PTEN-deficient cells. Interestingly, S6K2 exerted a dual role in signal transduction: it sustained GBM growth while also exerting negative feedback control on the upstream receptor tyrosine kinase AXL. Genetic inactivation of S6K2 sensitized PTEN-deficient GBMs to AXL inhibition, indicating that derepressed AXL is required for compensatory signaling upon targeting of the PI3K/Akt pathway. Combining the AXL inhibitor BMS-777607 with S6K1 inhibitor LY-2584702 also prevented compensatory AXL signaling and triggered cytotoxic responses selectively in the PTEN-deficient condition. Importantly, combination inhibition of AXL and S6K1 reduced pyrimidine biosynthesis, a known vulnerability of PTEN-deficient GBMs. The oral inhibitors of S6K1 and AXL were found to be brain-penetrant and effective in reducing GBM growth in several mouse models. These results establish that targeting of AXL can circumvent S6 Kinase dependent feedback to reduce pyrimidine biosynthesis and trigger cytotoxic responses in PTEN-deficient GBMs.

Calcineurin Inhibitors Synergize with Manogepix to Kill Diverse Human Fungal Pathogens.

Invasive fungal infections have mortality rates of 30–90%, depending on patient co-morbidities and the causative pathogen. The frequent emergence of drug resistance reduces the efficacy of currently approved treatment options, highlighting an urgent need for antifungals with new modes of action. Addressing this need, fosmanogepix (N-phosphonooxymethylene prodrug of manogepix; MGX) is the first in a new class of gepix drugs, and acts as a broad-spectrum, orally bioavailable inhibitor of the essential fungal glycosylphosphatidylinositol (GPI) acyltransferase Gwt1. MGX inhibits the growth of diverse fungal pathogens and causes accumulation of immature GPI-anchored proteins in the fungal endoplasmic reticulum. Relevant to the ongoing clinical development of fosmanogepix, we report a synergistic, fungicidal interaction between MGX and inhibitors of the protein phosphatase calcineurin against important human fungal pathogens. To investigate this synergy further, we evaluated a library of 124 conditional expression mutants covering 95% of the genes encoding proteins involved in GPI-anchor biosynthesis or proteins predicted to be GPI-anchored. Strong negative chemical-genetic interactions between the calcineurin inhibitor FK506 and eleven GPI-anchor biosynthesis genes were identified, indicating that calcineurin signalling is required for fungal tolerance to not only MGX, but to inhibition of the GPI-anchor biosynthesis pathway more broadly. Depletion of these GPI-anchor biosynthesis genes, like MGX treatment, also exposed fungal cell wall (1→3)-β-D-glucans. Taken together, these findings suggest the increased risk of invasive fungal infections associated with use of calcineurin inhibitors as immunosuppressants may be mitigated by their synergistic fungicidal interaction with (fos)manogepix and its ability to enhance exposure of immunostimulatory glucans.

BIONIC: biological network integration using convolutions.

Biological networks constructed from varied data can be used to map cellular function, but each data type has limitations. Network integration promises to address these limitations by combining and automatically weighting input information to obtain a more accurate and comprehensive representation of the underlying biology. We developed a deep learning-based network integration algorithm that incorporates a graph convolutional network framework. Our method, BIONIC (Biological Network Integration using Convolutions), learns features that contain substantially more functional information compared to existing approaches. BIONIC has unsupervised and semisupervised learning modes, making use of available gene function annotations. BIONIC is scalable in both size and quantity of the input networks, making it feasible to integrate numerous networks on the scale of the human genome. To demonstrate the use of BIONIC in identifying new biology, we predicted and experimentally validated essential gene chemical-genetic interactions from nonessential gene profiles in yeast.

Uncovering interactions between mycobacterial respiratory complexes to target drug-resistant Mycobacterium tuberculosis

Mycobacterium tuberculosis remains a leading cause of infectious disease morbidity and mortality for which new drug combination therapies are needed. Mycobacterial bioenergetics has emerged as a promising space for the development of novel therapeutics. Further to this, unique combinations of respiratory inhibitors have been shown to have synergistic or synthetic lethal interactions, suggesting that combinations of bioenergetic inhibitors could drastically shorten treatment times. Realizing the full potential of this unique target space requires an understanding of which combinations of respiratory complexes, when inhibited, have the strongest interactions and potential in a clinical setting. In this review, we discuss (i) chemical-interaction, (ii) genetic-interaction and (iii) chemical-genetic interaction studies to explore the consequences of inhibiting multiple mycobacterial respiratory components. We provide potential mechanisms to describe the basis for the strongest interactions. Finally, whilst we place an emphasis on interactions that occur with existing bioenergetic inhibitors, by highlighting interactions that occur with alternative respiratory components we envision that this information will provide a rational to further explore alternative proteins as potential drug targets and as part of unique drug combinations.

Chemical-genetic interaction mapping links carbon metabolism and cell wall structure to tuberculosis drug efficacy.

Current chemotherapy against Mycobacterium tuberculosis (Mtb), an important human pathogen, requires a multidrug regimen lasting several months. While efforts have been made to optimize therapy by exploiting drug–drug synergies, testing new drug combinations in relevant host environments remains arduous. In particular, host environments profoundly affect the bacterial metabolic state and drug efficacy, limiting the accuracy of predictions based on in vitro assays alone. In this study, we utilized conditional Mtb knockdown mutants of essential genes as an experimentally tractable surrogate for drug treatment and probe the relationship between Mtb carbon metabolism and chemical–genetic interactions (CGIs). We examined the antitubercular drugs isoniazid, rifampicin, and moxifloxacin and found that CGIs are differentially responsive to the metabolic state, defining both environment-independent and -dependent interactions. Specifically, growth on the in vivo–relevant carbon source, cholesterol, reduced rifampicin efficacy by altering mycobacterial cell surface lipid composition. We report that a variety of perturbations in cell wall synthesis pathways restore rifampicin efficacy during growth on cholesterol, and that both environment-independent and cholesterol-dependent in vitro CGIs could be leveraged to enhance bacterial clearance in the mouse infection model. Our findings present an atlas of chemical–genetic–environmental interactions that can be used to optimize drug–drug interactions, as well as provide a framework for understanding in vitro correlates of in vivo efficacy.

Deep learning-driven prediction of drug mechanism of action from large-scale chemical-genetic interaction profiles

MotivationChemical–genetic interaction profiling is a genetic approach that quantifies the susceptibility of a set of mutants depleted in specific gene product(s) to a set of chemical compounds. With the recent advances in artificial intelligence, chemical–genetic interaction profiles (CGIPs) can be leveraged to predict mechanism of action of compounds. This can be achieved by using machine learning, where the data from a CGIP is fed into the machine learning platform along with the chemical descriptors to develop a chemogenetically trained model. As small molecules can be considered non-structural data, graph convolutional neural networks, which can learn from the chemical structures directly, can be used to successfully predict molecular properties. Clustering analysis, on the other hand, is a critical approach to get insights into the underlying biological relationships between the gene products in the high-dimensional chemical-genetic data.Methods and resultsIn this study, we proposed a comprehensive framework based on the large-scale chemical-genetics dataset built in Mycobacterium tuberculosis for predicting CGIPs using graph-based deep learning models. Our approach is structured into three parts. First, by matching M. tuberculosis genes with homologous genes in Escherichia coli (E. coli) according to their gene products, we grouped the genes into clusters with distinct biological functions. Second, we employed a directed message passing neural network to predict growth inhibition against M. tuberculosis gene clusters using a collection of 50,000 chemicals with the profile. We compared the performance of different baseline models and implemented multi-label tasks in binary classification frameworks. Lastly, we applied the trained model to an externally curated drug set that had experimental results against M. tuberculosis genes to examine the effectiveness of our method. Overall, we demonstrate that our approach effectively created M. tuberculosis gene clusters, and the trained classifier is able to predict activity against essential M. tuberculosis targets with high accuracy.ConclusionThis work provides an analytical framework for modeling large-scale chemical-genetic datasets for predicting CGIPs and generating hypothesis about mechanism of action of novel drugs. In addition, this work highlights the importance of graph-based deep neural networks in drug discovery.