AI for targeted polypharmacology: The next frontier in drug discovery.

In drug discovery, targeted polypharmacology, i.e., targeting multiple molecular targets with a single drug, is redefining therapeutic design to address complex diseases. Pre-selected pharmacological profiles, as exemplified in kinase drugs, promise enhanced efficacy and reduced toxicity. Historically, many of such drugs were discovered serendipitously, limiting predictability and efficacy, but currently artificial intelligence (AI) offers a transformative solution. Machine learning and deep learning techniques enable modeling protein structures, generating novel compounds, and decoding their polypharmacological effects, opening an avenue for more systematic and predictive multi-target drug design. This review explores the use of AI in identifying synergistic co-targets and delineating them from anti-targets that lead to adverse effects, and then discusses advances in AI-enabled docking, generative chemistry, and proteochemometric modeling of proteome-wide compound interactions, in the context of polypharmacology. We also provide insights into challenges ahead.

Investigating the conformational landscape of AlphaFold2-predicted protein kinase structures.

Summary: Protein kinases are a family of signaling proteins, crucial for maintaining cellular homeostasis. When dysregulated, kinases drive the pathogenesis of several diseases, and are thus one of the largest target categories for drug discovery. Kinase activity is tightly controlled by switching through several active and inactive conformations in their catalytic domain. Kinase inhibitors have been designed to engage kinases in specific conformational states, where each conformation presents a unique physico-chemical environment for therapeutic intervention. Thus, modeling kinases across conformations can enable the design of novel and optimally selective kinase drugs. Due to the recent success of AlphaFold2 in accurately predicting the 3D structure of proteins based on sequence, we investigated the conformational landscape of protein kinases as modeled by AlphaFold2. We observed that AlphaFold2 is able to model several kinase conformations across the kinome, however, certain conformations are only observed in specific kinase families. Furthermore, we show that the per residue predicted local distance difference test can capture information describing structural flexibility of kinases. Finally, we evaluated the docking performance of AlphaFold2 kinase structures for enriching known ligands. Taken together, we see an opportunity to leverage AlphaFold2 models for structure-based drug discovery against kinases across several pharmacologically relevant conformational states. Availability and implementation: All code used in the analysis is freely available at https://github.com/Harmonic-Discovery/AF2-kinase-conformational-landscape.

Structural signatures: a web server for exploring a database of and generating protein structural features from human cell lines and tissues.

Structural features of proteins provide powerful insights into biological function and similarity. Specifically, previous work has demonstrated that structural features of tissue and drug-treated cell line samples can be used to predict tissue type and characterize drug relationships, respectively. We have developed structural signatures, a web server for annotating and analyzing protein features from gene sets that are often found in transcriptomic and proteomic data. This platform provides access to a structural feature database derived from normal and disease human tissue samples. We show how analysis using this database can shed light on the relationship between states of single-cell RNA-sequencing lung cancer samples. These various structural feature signatures can be visualized on the server itself or downloaded for additional analysis. The structural signatures server tool is freely available at https://structural-server.kinametrix.com/.

Protein structure-based gene expression signatures.

Gene expression signatures (GES) connect phenotypes to differential messenger RNA (mRNA) expression of genes, providing a powerful approach to define cellular identity, function, and the effects of perturbations. The use of GES has suffered from vague assessment criteria and limited reproducibility. Because the structure of proteins defines the functional capability of genes, we hypothesized that enrichment of structural features could be a generalizable representation of gene sets. We derive structural gene expression signatures (sGES) using features from multiple levels of protein structure (e.g., domain and fold) encoded by the mRNAs in GES. Comprehensive analyses of data from the Genotype-Tissue Expression Project (GTEx), the all RNA-seq and ChIP-seq sample and signature search (ARCHS4) database, and mRNA expression of drug effects on cardiomyocytes show that sGES are useful for characterizing biological phenomena. sGES enable phenotypic characterization across experimental platforms, facilitates interoperability of expression datasets, and describe drug action on cells.

Maternal health around pregnancy and autism risk: a diagnosis-wide, population-based study.

BACKGROUND: Many studies have reported an increased risk of autism spectrum disorder (ASD) associated with some maternal diagnoses in pregnancy. However, such associations have not been studied systematically, accounting for comorbidity between maternal disorders. Therefore our aim was to comprehensively test the associations between maternal diagnoses around pregnancy and ASD risk in offspring. METHODS: This exploratory case-cohort study included children born in Israel from 1997 to 2008, and followed up until 2015. We used information on all ICD-9 codes received by their mothers during pregnancy and the preceding year. ASD risk associated with each of those conditions was calculated using Cox proportional hazards regression, adjusted for the confounders (birth year, maternal age, socioeconomic status and number of ICD-9 diagnoses during the exposure period). RESULTS: The analytic sample consisted of 80 187 individuals (1132 cases, 79 055 controls), with 822 unique ICD-9 codes recorded in their mothers. After extensive quality control, 22 maternal diagnoses were nominally significantly associated with offspring ASD, with 16 of those surviving subsequent filtering steps (permutation testing, multiple testing correction, multiple regression). Among those, we recorded an increased risk of ASD associated with metabolic [e.g. hypertension; HR = 2.74 (1.92-3.90), p = 2.43 × 10-8], genitourinary [e.g. non-inflammatory disorders of cervix; HR = 1.88 (1.38-2.57), p = 7.06 × 10-5] and psychiatric [depressive disorder; HR = 2.11 (1.32-3.35), p = 1.70 × 10-3] diagnoses. Meanwhile, mothers of children with ASD were less likely to attend prenatal care appointment [HR = 0.62 (0.54-0.71), p = 1.80 × 10-11]. CONCLUSIONS: Sixteen maternal diagnoses were associated with ASD in the offspring, after rigorous filtering of potential false-positive associations. Replication in other cohorts and further research to understand the mechanisms underlying the observed associations with ASD are warranted.

Transcriptomic profiling of human cardiac cells predicts protein kinase inhibitor-associated cardiotoxicity.

Kinase inhibitors (KIs) represent an important class of anti-cancer drugs. Although cardiotoxicity is a serious adverse event associated with several KIs, the reasons remain poorly understood, and its prediction remains challenging. We obtain transcriptional profiles of human heart-derived primary cardiomyocyte like cell lines treated with a panel of 26 FDA-approved KIs and classify their effects on subcellular pathways and processes. Individual cardiotoxicity patient reports for these KIs, obtained from the FDA Adverse Event Reporting System, are used to compute relative risk scores. These are then combined with the cell line-derived transcriptomic datasets through elastic net regression analysis to identify a gene signature that can predict risk of cardiotoxicity. We also identify relationships between cardiotoxicity risk and structural/binding profiles of individual KIs. We conclude that acute transcriptomic changes in cell-based assays combined with drug substructures are predictive of KI-induced cardiotoxicity risk, and that they can be informative for future drug discovery.

Identification of newborns at risk for autism using electronic medical records and machine learning.

BACKGROUND: Current approaches for early identification of individuals at high risk for autism spectrum disorder (ASD) in the general population are limited, and most ASD patients are not identified until after the age of 4. This is despite substantial evidence suggesting that early diagnosis and intervention improves developmental course and outcome. The aim of the current study was to test the ability of machine learning (ML) models applied to electronic medical records (EMRs) to predict ASD early in life, in a general population sample. METHODS: We used EMR data from a single Israeli Health Maintenance Organization, including EMR information for parents of 1,397 ASD children (ICD-9/10) and 94,741 non-ASD children born between January 1st, 1997 and December 31st, 2008. Routinely available parental sociodemographic information, parental medical histories, and prescribed medications data were used to generate features to train various ML algorithms, including multivariate logistic regression, artificial neural networks, and random forest. Prediction performance was evaluated with 10-fold cross-validation by computing the area under the receiver operating characteristic curve (AUC; C-statistic), sensitivity, specificity, accuracy, false positive rate, and precision (positive predictive value [PPV]). RESULTS: All ML models tested had similar performance. The average performance across all models had C-statistic of 0.709, sensitivity of 29.93%, specificity of 98.18%, accuracy of 95.62%, false positive rate of 1.81%, and PPV of 43.35% for predicting ASD in this dataset. CONCLUSIONS: We conclude that ML algorithms combined with EMR capture early life ASD risk as well as reveal previously unknown features to be associated with ASD-risk. Such approaches may be able to enhance the ability for accurate and efficient early detection of ASD in large populations of children.

Systems-level analysis of NalD mutation, a recurrent driver of rapid drug resistance in acute Pseudomonas aeruginosa infection.

Pseudomonas aeruginosa, a main cause of human infection, can gain resistance to the antibiotic aztreonam through a mutation in NalD, a transcriptional repressor of cellular efflux. Here we combine computational analysis of clinical isolates, transcriptomics, metabolic modeling and experimental validation to find a strong association between NalD mutations and resistance to aztreonam-as well as resistance to other antibiotics-across P. aeruginosa isolated from different patients. A detailed analysis of one patient's timeline shows how this mutation can emerge in vivo and drive rapid evolution of resistance while the patient received cancer treatment, a bone marrow transplantation, and antibiotics up to the point of causing the patient's death. Transcriptomics analysis confirmed the primary mechanism of NalD action-a loss-of-function mutation that caused constitutive overexpression of the MexAB-OprM efflux system-which lead to aztreonam resistance but, surprisingly, had no fitness cost in the absence of the antibiotic. We constrained a genome-scale metabolic model using the transcriptomics data to investigate changes beyond the primary mechanism of resistance, including adaptations in major metabolic pathways and membrane transport concurrent with aztreonam resistance, which may explain the lack of a fitness cost. We propose that metabolic adaptations may allow resistance mutations to endure in the absence of antibiotics and could be targeted by future therapies against antibiotic resistant pathogens.

KinaMetrix: a web resource to investigate kinase conformations and inhibitor space.

Protein kinases are among the most explored protein drug targets. Visualization of kinase conformations is critical for understanding structure-function relationship in this family and for developing chemically unique, conformation-specific small molecule drugs. We have developed Kinformation, a random forest classifier that annotates the conformation of over 3500 protein kinase structures in the Protein Data Bank. Kinformation was trained on structural descriptors derived from functionally important motifs to automatically categorize kinases into five major conformations with pharmacological relevance. Here we present KinaMetrix (http://KinaMetrix.com), a web resource enabling researchers to investigate the protein kinase conformational space as well as a subset of kinase inhibitors that exhibit conformational specificity. KinaMetrix allows users to classify uploaded kinase structures, as well as to derive structural descriptors of protein kinases. Uploaded structures can then be compared to atomic structures of other kinases, enabling users to identify kinases that occupy a similar conformational space to their uploaded structure. Finally, KinaMetrix also serves as a repository for both small molecule substructures that are significantly associated with each conformation type, and for homology models of kinases in inactive conformations. We expect KinaMetrix to serve as a resource for researchers studying kinase structural biology or developing conformation-specific kinase inhibitors.

Association of Autism Spectrum Disorder With Prenatal Exposure to Medication Affecting Neurotransmitter Systems.

Importance: Prenatal exposure to certain medications has been hypothesized to influence the risk of autism spectrum disorders (ASD). However, the underlying effects on the neurotransmitter systems have not been comprehensively assessed. Objective: To investigate the association of early-life interference with different neurotransmitter systems by prenatal medication exposure on the risk of ASD in offspring. Design, Setting, and Participants: This case-control study included children born from January 1, 1997, through December 31, 2007, and followed up for ASD until January 26, 2015, within a single Israeli health maintenance organization. Using publicly available data, 55 groups of medications affecting neurotransmitter systems and prescribed to pregnant women in this sample were identified. Children prenatally exposed to medications were compared with nonexposed children. Data were analyzed from March 1, 2017, through June 20, 2018. Main Outcome and Measures: Hazard ratios (HRs) and 95% CIs of ASD risk associated with exposure to medication groups using Cox proportional hazards regression, adjusted for the relevant confounders (eg, birth year, maternal age, maternal history of psychiatric and neurologic disorders, or maternal number of all medical diagnoses 1 year before pregnancy). Results: The analytic sample consisted of 96 249 individuals (1405 cases; 94 844 controls; mean [SD] age at the end of follow-up, 11.6 [3.1] years; 48.8% female), including 1405 with ASD and 94 844 controls. Of 34 groups of medications, 5 showed nominally statistically significant association with ASD in fully adjusted models. Evidence of confounding effects of the number of maternal diagnoses on the association between offspring exposure to medication and ASD was found. Adjusting for this factor, lower estimates of ASD risk among children exposed to cannabinoid receptor agonists (HR, 0.72; 95% CI, 0.55-0.95; P = .02), muscarinic receptor 2 agonists (HR, 0.49; 95% CI, 0.24-0.98; P = .04), opioid receptor κ and ε agonists (HR, 0.67; 95% CI, 0.45-0.99; P = .045), or α2C-adrenergic receptor agonists (HR, 0.43; 95% CI, 0.19-0.96; P = .04) were observed. Exposure to antagonists of neuronal nicotinic acetylcholine receptor α was associated with higher estimates of ASD risk (HR, 12.94; 95% CI, 1.35-124.25; P = .03). Conclusions and Relevance: Most of the medications affecting neurotransmitter systems in this sample had no association with the estimates of ASD risk. Replication and/or validation using experimental techniques are required.

Redefining the Protein Kinase Conformational Space with Machine Learning.

Protein kinases are dynamic, adopting different conformational states that are critical for their catalytic activity. We assess a range of structural features derived from the conserved αC helix and DFG motif to define the conformational space of the catalytic domain of protein kinases. We then construct Kinformation, a random forest classifier, to annotate the conformation of 3,708 kinase structures in the PDB. Our classification scheme captures known active and inactive kinase conformations and defines an additional conformational state, thereby refining the current understanding of the kinase conformational space. Furthermore, network analysis of the small molecules recognized by each conformation captures chemical substructures that are associated with each conformation type. Our description of the kinase conformational space is expected to improve modeling of protein kinase structures, as well as guide the development of conformation-specific kinase inhibitors with optimal pharmacological profiles.

Correction: Generation of G protein-coupled receptor antibodies differentially sensitive to conformational states.

[This corrects the article DOI: 10.1371/journal.pone.0187306.].

Bow-tie signaling in c-di-GMP: Machine learning in a simple biochemical network.

Bacteria of many species rely on a simple molecule, the intracellular secondary messenger c-di-GMP (Bis-(3'-5')-cyclic dimeric guanosine monophosphate), to make a vital choice: whether to stay in one place and form a biofilm, or to leave it in search of better conditions. The c-di-GMP network has a bow-tie shaped architecture that integrates many signals from the outside world-the input stimuli-into intracellular c-di-GMP levels that then regulate genes for biofilm formation or for swarming motility-the output phenotypes. How does the 'uninformed' process of evolution produce a network with the right input/output association and enable bacteria to make the right choice? Inspired by new data from 28 clinical isolates of Pseudomonas aeruginosa and strains evolved in laboratory experiments we propose a mathematical model where the c-di-GMP network is analogous to a machine learning classifier. The analogy immediately suggests a mechanism for learning through evolution: adaptation though incremental changes in c-di-GMP network proteins acquires knowledge from past experiences and enables bacteria to use it to direct future behaviors. Our model clarifies the elusive function of the ubiquitous c-di-GMP network, a key regulator of bacterial social traits associated with virulence. More broadly, the link between evolution and machine learning can help explain how natural selection across fluctuating environments produces networks that enable living organisms to make sophisticated decisions.