Large-Scale Screening of Per- and Polyfluoroalkyl Substance Binding Interactions and Their Mixtures with Nuclear Receptors.

Despite their significant impact, comprehensive screenings and detailed analyses of per- and polyfluoroalkyl substance (PFAS) binding strengths at the orthosteric and allosteric sites of NRs are currently lacking. This study addresses this gap by focusing on the binding interaction analysis of both common and uncommon PFAS with the nuclear receptors (NRs) vitamin D receptor (VDR), peroxisome proliferator-activated receptor gamma (PPARγ), pregnane X receptor (PXR), and estrogen receptor alpha (ERα). Advanced docking simulations were used to screen 9507 PFAS chemicals at the orthosteric and allosteric sites of PPARγ, PXR, VDR, and ERα. All receptors exhibited strong binding interactions at the orthosteric and allosteric site with a significant number of PFAS. We verified the accuracy of the docking protocol through multiple docking controls and validations. A mixture modeling analysis indicates that PFAS can bind in various combinations with themselves and endogenous ligands simultaneously, to disrupt the endocrine system and cause carcinogenic responses. These findings reveal that PFAS can interfere with nuclear receptor activity by displacing endogenous or native ligands by binding to the orthosteric and allosteric sites. The purpose of this study is to explore the mechanisms through which PFAS exert their endocrine-disrupting effects, potentially leading to more targeted therapeutic strategies. Importantly, this study is the first to explore the binding of PFAS at allosteric sites and to model PFAS mixtures at nuclear receptors. Given the high concentration and persistence of PFAS in humans, this study further emphasizes the urgent need for further research into the carcinogenic mechanisms of PFAS and the development of therapeutic strategies that target nuclear receptors.

Machine and deep learning approaches for cancer drug repurposing.

Knowledge of the underpinnings of cancer initiation, progression and metastasis has increased exponentially in recent years. Advanced "omics" coupled with machine learning and artificial intelligence (deep learning) methods have helped elucidate targets and pathways critical to those processes that may be amenable to pharmacologic modulation. However, the current anti-cancer therapeutic armamentarium continues to lag behind. As the cost of developing a new drug remains prohibitively expensive, repurposing of existing approved and investigational drugs is sought after given known safety profiles and reduction in the cost barrier. Notably, successes in oncologic drug repurposing have been infrequent. Computational in-silico strategies have been developed to aid in modeling biological processes to find new disease-relevant targets and discovering novel drug-target and drug-phenotype associations. Machine and deep learning methods have especially enabled leaps in those successes. This review will discuss these methods as they pertain to cancer biology as well as immunomodulation for drug repurposing opportunities in oncologic diseases.

Cadherin 11 Promotes Immunosuppression and Extracellular Matrix Deposition to Support Growth of Pancreatic Tumors and Resistance to Gemcitabine in Mice.

BACKGROUND & AIMS: Pancreatic ductal adenocarcinomas (PDACs) are characterized by fibrosis and an abundance of cancer-associated fibroblasts (CAFs). We investigated strategies to disrupt interactions among CAFs, the immune system, and cancer cells, focusing on adhesion molecule CDH11, which has been associated with other fibrotic disorders and is expressed by activated fibroblasts. METHODS: We compared levels of CDH11 messenger RNA in human pancreatitis and pancreatic cancer tissues and cells with normal pancreas, and measured levels of CDH11 protein in human and mouse pancreatic lesions and normal tissues. We crossed p48-Cre;LSL-KrasG12D/+;LSL-Trp53R172H/+ (KPC) mice with CDH11-knockout mice and measured survival times of offspring. Pancreata were collected and analyzed by histology, immunohistochemistry, and (single-cell) RNA sequencing; RNA and proteins were identified by imaging mass cytometry. Some mice were given injections of PD1 antibody or gemcitabine and survival was monitored. Pancreatic cancer cells from KPC mice were subcutaneously injected into Cdh11+/+ and Cdh11-/- mice and tumor growth was monitored. Pancreatic cancer cells (mT3) from KPC mice (C57BL/6), were subcutaneously injected into Cdh11+/+ (C57BL/6J) mice and mice were given injections of antibody against CDH11, gemcitabine, or small molecule inhibitor of CDH11 (SD133) and tumor growth was monitored. RESULTS: Levels of CDH11 messenger RNA and protein were significantly higher in CAFs than in pancreatic cancer epithelial cells, human or mouse pancreatic cancer cell lines, or immune cells. KPC/Cdh11+/- and KPC/Cdh11-/- mice survived significantly longer than KPC/Cdh11+/+ mice. Markers of stromal activation entirely surrounded pancreatic intraepithelial neoplasias in KPC/Cdh11+/+ mice and incompletely in KPC/Cdh11+/- and KPC/Cdh11-/- mice, whose lesions also contained fewer FOXP3+ cells in the tumor center. Compared with pancreatic tumors in KPC/Cdh11+/+ mice, tumors of KPC/Cdh11+/- mice had increased markers of antigen processing and presentation; more lymphocytes and associated cytokines; decreased extracellular matrix components; and reductions in markers and cytokines associated with immunosuppression. Administration of the PD1 antibody did not prolong survival of KPC mice with 0, 1, or 2 alleles of Cdh11. Gemcitabine extended survival of KPC/Cdh11+/- and KPC/Cdh11-/- mice only or reduced subcutaneous tumor growth in mT3 engrafted Cdh11+/+ mice when given in combination with the CDH11 antibody. A small molecule inhibitor of CDH11 reduced growth of pre-established mT3 subcutaneous tumors only if T and B cells were present in mice. CONCLUSIONS: Knockout or inhibition of CDH11, which is expressed by CAFs in the pancreatic tumor stroma, reduces growth of pancreatic tumors, increases their response to gemcitabine, and significantly extends survival of mice. CDH11 promotes immunosuppression and extracellular matrix deposition, and might be developed as a therapeutic target for pancreatic cancer.

A novel chemo-phenotypic method identifies mixtures of salpn, vitamin D3, and pesticides involved in the development of colorectal and pancreatic cancer.

Environmental chemical (EC) exposures and our interactions with them has significantly increased in the recent decades. Toxicity associated biological characterization of these chemicals is challenging and inefficient, even with available high-throughput technologies. In this report, we describe a novel computational method for characterizing toxicity, associated biological perturbations and disease outcome, called the Chemo-Phenotypic Based Toxicity Measurement (CPTM). CPTM is used to quantify the EC "toxicity score" (Zts), which serves as a holistic metric of potential toxicity and disease outcome. CPTM quantitative toxicity is the measure of chemical features, biological phenotypic effects, and toxicokinetic properties of the ECs. For proof-of-concept, we subject ECs obtained from the Environmental Protection Agency's (EPA) database to the CPTM. We validated the CPTM toxicity predictions by correlating 'Zts' scores with known toxicity effects. We also confirmed the CPTM predictions with in-vitro, and in-vivo experiments. In in-vitro and zebrafish models, we showed that, mixtures of the motor oil and food additive 'Salpn' with endogenous nuclear receptor ligands such as Vitamin D3, dysregulated the nuclear receptors and key transcription pathways involved in Colorectal Cancer. Further, in a human patient derived cell organoid model, we found that a mixture of the widely used pesticides 'Tetramethrin' and 'Fenpropathrin' significantly impacts the population of patient derived pancreatic cancer cells and 3D organoid models to support rapid PDAC disease progression. The CPTM method is, to our knowledge, the first comprehensive toxico-physicochemical, and phenotypic bionetwork-based platform for efficient high-throughput screening of environmental chemical toxicity, mechanisms of action, and connection to disease outcomes.

Predicting Chemical Carcinogens Using a Hybrid Neural Network Deep Learning Method.

Determining environmental chemical carcinogenicity is urgently needed as humans are increasingly exposed to these chemicals. In this study, we developed a hybrid neural network (HNN) method called HNN-Cancer to predict potential carcinogens of real-life chemicals. The HNN-Cancer included a new SMILES feature representation method by modifying our previous 3D array representation of 1D SMILES simulated by the convolutional neural network (CNN). We developed binary classification, multiclass classification, and regression models based on diverse non-congeneric chemicals. Along with the HNN-Cancer model, we developed models based on the random forest (RF), bootstrap aggregating (Bagging), and adaptive boosting (AdaBoost) methods for binary and multiclass classification. We developed regression models using HNN-Cancer, RF, support vector regressor (SVR), gradient boosting (GB), kernel ridge (KR), decision tree with AdaBoost (DT), KNeighbors (KN), and a consensus method. The performance of the models for all classifications was assessed using various statistical metrics. The accuracy of the HNN-Cancer, RF, and Bagging models were 74%, and their AUC was ~0.81 for binary classification models developed with 7994 chemicals. The sensitivity was 79.5% and the specificity was 67.3% for the HNN-Cancer, which outperforms the other methods. In the case of multiclass classification models with 1618 chemicals, we obtained the optimal accuracy of 70% with an AUC 0.7 for HNN-Cancer, RF, Bagging, and AdaBoost, respectively. In the case of regression models, the correlation coefficient (R) was around 0.62 for HNN-Cancer and RF higher than the SVM, GB, KR, DTBoost, and NN machine learning methods. Overall, the HNN-Cancer performed better for the majority of the known carcinogen experimental datasets. Further, the predictive performance of HNN-Cancer on diverse chemicals is comparable to the literature-reported models that included similar and less diverse molecules. Our HNN-Cancer could be used in identifying potentially carcinogenic chemicals for a wide variety of chemical classes.

A New Hybrid Neural Network Deep Learning Method for Protein-Ligand Binding Affinity Prediction and De Novo Drug Design.

Accurately predicting ligand binding affinity in a virtual screening campaign is still challenging. Here, we developed hybrid neural network (HNN) machine deep learning methods, HNN-denovo and HNN-affinity, by combining the 3D-CNN (convolutional neural network) and the FFNN (fast forward neural network) hybrid neural network framework. The HNN-denovo uses protein pocket structure and protein-ligand interactions as input features. The HNN-affinity uses protein sequences and ligand features as input features. The HNN method combines the CNN and FCNN machine architecture for the protein structure or protein sequence and ligand descriptors. To train the model, the HNN methods used thousands of known protein-ligand binding affinity data retrieved from the PDBBind database. We also developed the Random Forest (RF), Gradient Boosting (GB), Decision Tree with AdaBoost (DT), and a consensus model. We compared the HNN results with models developed based on the RF, GB, and DT methods. We also independently compared the HNN method results with the literature reported deep learning protein-ligand binding affinity predictions made by the DLSCORE, KDEEP, and DeepAtom. The predictive performance of the HNN methods (max Pearson's R achieved was 0.86) was consistently better than or comparable to the DLSCORE, KDEEP, and DeepAtom deep learning learning methods for both balanced and unbalanced data sets. The HNN-affinity can be applied for the protein-ligand affinity prediction even in the absence of protein structure information, as it considers the protein sequence as standalone feature in addition to the ligand descriptors. The HNN-denovo method can be efficiently implemented to the structure-based de novo drug design campaign. The HNN-affinity method can be used in conjunction with the deep learning molecular docking protocols as a standalone. Further, it can be combined with the conventional molecular docking methods as a multistep approach to rapidly screen billions of diverse compounds. The HNN method are highly scalable in the cloud ML platform.

ES-Screen: A Novel Electrostatics-Driven Method for Drug Discovery Virtual Screening.

Electrostatic interactions drive biomolecular interactions and associations. Computational modeling of electrostatics in biomolecular systems, such as protein-ligand, protein-protein, and protein-DNA, has provided atomistic insights into the binding process. In drug discovery, finding biologically plausible ligand-protein target interactions is challenging as current virtual screening and adjuvant techniques such as docking methods do not provide optimal treatment of electrostatic interactions. This study describes a novel electrostatics-driven virtual screening method called 'ES-Screen' that performs well across diverse protein target systems. ES-Screen provides a unique treatment of electrostatic interaction energies independent of total electrostatic free energy, typically employed by current software. Importantly, ES-Screen uses initial ligand pose input obtained from a receptor-based pharmacophore, thus independent of molecular docking. ES-Screen integrates individual polar and nonpolar replacement energies, which are the energy costs of replacing the cognate ligand for a target with a query ligand from the screening. This uniquely optimizes thermodynamic stability in electrostatic and nonpolar interactions relative to an experimentally determined stable binding state. ES-Screen also integrates chemometrics through shape and other physicochemical properties to prioritize query ligands with the greatest physicochemical similarities to the cognate ligand. The applicability of ES-Screen is demonstrated with in vitro experiments by identifying novel targets for many drugs. The present version includes a combination of many other descriptor components that, in a future version, will be purely based on electrostatics. Therefore, ES-Screen is a first-in-class unique electrostatics-driven virtual screening method with a unique implementation of replacement electrostatic interaction energies with broad applicability in drug discovery.

Proglumide Reverses Nonalcoholic Steatohepatitis by Interaction with the Farnesoid X Receptor and Altering the Microbiome.

Nonalcoholic steatohepatitis (NASH) is associated with obesity, metabolic syndrome, and dysbiosis of the gut microbiome. Cholecystokinin (CCK) is released by saturated fats and plays an important role in bile acid secretion. CCK receptors are expressed on cholangiocytes, and CCK-B receptor expression increases in the livers of mice with NASH. The farnesoid X receptor (FXR) is involved in bile acid transport and is a target for novel therapeutics for NASH. The aim of this study was to examine the role of proglumide, a CCK receptor inhibitor, in a murine model of NASH and its interaction at FXR. Mice were fed a choline deficient ethionine (CDE) diet to induce NASH. Some CDE-fed mice received proglumide-treated drinking water. Blood was collected and liver tissues were examined histologically. Proglumide's interaction at FXR was evaluated by computer modeling, a luciferase reporter assay, and tissue FXR expression. Stool microbiome was analyzed by RNA-Sequencing. CDE-fed mice developed NASH and the effect was prevented by proglumide. Computer modeling demonstrated specific binding of proglumide to FXR. Proglumide binding in the reporter assay was consistent with a partial agonist at the FXR with a mean binding affinity of 215 nM. FXR expression was significantly decreased in livers of CDE-fed mice compared to control livers, and proglumide restored FXR expression to normal levels. Proglumide therapy altered the microbiome signature by increasing beneficial and decreasing harmful bacteria. These data highlight the potential novel mechanisms by which proglumide therapy may improve NASH through interaction with the FXR and consequent alteration of the gut microbiome.

Predicting new drug indications for prostate cancer: The integration of an in silico proteochemometric network pharmacology platform with patient-derived primary prostate cells.

BACKGROUND: Drug repurposing enables the discovery of potential cancer treatments using publically available data from over 4000 published Food and Drug Administration approved and experimental drugs. However, the ability to effectively evaluate the drug's efficacy remains a challenge. Impediments to broad applicability include inaccuracies in many of the computational drug-target algorithms and a lack of clinically relevant biologic modeling systems to validate the computational data for subsequent translation. METHODS: We have integrated our computational proteochemometric systems network pharmacology platform, DrugGenEx-Net, with primary, continuous cultures of conditionally reprogrammed (CR) normal and prostate cancer (PCa) cells derived from treatment-naive patients with primary PCa. RESULTS: Using the transcriptomic data from two matched pairs of benign and tumor-derived CR cells, we constructed drug networks to describe the biological perturbation associated with each prostate cell subtype at multiple levels of biological action. We prioritized the drugs by analyzing these networks for statistical coincidence with the drug action networks originating from known and predicted drug-protein targets. Prioritized drugs shared between the two patients' PCa cells included carfilzomib (CFZ), bortezomib (BTZ), sulforaphane, and phenethyl isothiocyanate. The effects of these compounds were then tested in the CR cells, in vitro. We observed that the IC50 values of the normal PCa CR cells for CFZ and BTZ were higher than their matched tumor CR cells. Transcriptomic analysis of CFZ-treated CR cells revealed that genes involved in cell proliferation, proteases, and downstream targets of serine proteases were inhibited while KLK7 and KLK8 were induced in the tumor-derived CR cells. CONCLUSIONS: Given that the drugs in the database are extremely well-characterized and that the patient-derived cells are easily scalable for high throughput drug screening, this combined in vitro and in silico approach may significantly advance personalized PCa treatment and for other cancer applications.

DrugGenEx-Net: a novel computational platform for systems pharmacology and gene expression-based drug repurposing.

BACKGROUND: The targeting of disease-related proteins is important for drug discovery, and yet target-based discovery has not been fruitful. Contextualizing overall biological processes is critical to formulating successful drug-disease hypotheses. Network pharmacology helps to overcome target-based bottlenecks through systems biology analytics, such as protein-protein interaction (PPI) networks and pathway regulation. RESULTS: We present a systems polypharmacology platform entitled DrugGenEx-Net (DGE-NET). DGE-NET predicts empirical drug-target (DT) interactions, integrates interaction pairs into a multi-tiered network analysis, and ultimately predicts disease-specific drug polypharmacology through systems-based gene expression analysis. Incorporation of established biological network annotations for protein target-disease, -signaling pathway, -molecular function, and protein-protein interactions enhances predicted DT effects on disease pathophysiology. Over 50 drug-disease and 100 drug-pathway predictions are validated. For example, the predicted systems pharmacology of the cholesterol-lowering agent ezetimibe corroborates its potential carcinogenicity. When disease-specific gene expression analysis is integrated, DGE-NET prioritizes known therapeutics/experimental drugs as well as their contra-indications. Proof-of-concept is established for immune-related rheumatoid arthritis and inflammatory bowel disease, as well as neuro-degenerative Alzheimer's and Parkinson's diseases. CONCLUSIONS: DGE-NET is a novel computational method that predicting drug therapeutic and counter-therapeutic indications by uniquely integrating systems pharmacology with gene expression analysis. DGE-NET correctly predicts various drug-disease indications by linking the biological activity of drugs and diseases at multiple tiers of biological action, and is therefore a useful approach to identifying drug candidates for re-purposing.

Germ line variants of human N-methylpurine DNA glycosylase show impaired DNA repair activity and facilitate 1,N6-ethenoadenine-induced mutations.

Human N-methylpurine DNA glycosylase (hMPG) initiates base excision repair of a number of structurally diverse purine bases including 1,N(6)-ethenoadenine, hypoxanthine, and alkylation adducts in DNA. Genetic studies discovered at least eight validated non-synonymous single nucleotide polymorphisms (nsSNPs) of the hMPG gene in human populations that result in specific single amino acid substitutions. In this study, we tested the functional consequences of these nsSNPs of hMPG. Our results showed that two specific arginine residues, Arg-141 and Arg-120, are important for the activity of hMPG as the germ line variants R120C and R141Q had reduced enzymatic activity in vitro as well as in mammalian cells. Expression of these two variants in mammalian cells lacking endogenous MPG also showed an increase in mutations and sensitivity to an alkylating agent compared with the WT hMPG. Real time binding experiments by surface plasmon resonance spectroscopy suggested that these variants have substantial reduction in the equilibrium dissociation constant of binding (KD) of hMPG toward 1,N(6)-ethenoadenine-containing oligonucleotide (ϵA-DNA). Pre-steady-state kinetic studies showed that the substitutions at arginine residues affected the turnover of the enzyme significantly under multiple turnover condition. Surface plasmon resonance spectroscopy further showed that both variants had significantly decreased nonspecific (undamaged) DNA binding. Molecular modeling suggested that R141Q substitution may have resulted in a direct loss of the salt bridge between ϵA-DNA and hMPG, whereas R120C substitution redistributed, at a distance, the interactions among residues in the catalytic pocket. Together our results suggest that individuals carrying R120C and R141Q MPG variants may be at risk for genomic instability and associated diseases as a consequence.

Naturally occurring polyphenol, morin hydrate, inhibits enzymatic activity of N-methylpurine DNA glycosylase, a DNA repair enzyme with various roles in human disease.

Interest in the mechanisms of DNA repair pathways, including the base excision repair (BER) pathway specifically, has heightened since these pathways have been shown to modulate important aspects of human disease. Modulation of the expression or activity of a particular BER enzyme, N-methylpurine DNA glycosylase (MPG), has been demonstrated to play a role in carcinogenesis and resistance to chemotherapy as well as neurodegenerative diseases, which has intensified the focus on studying MPG-related mechanisms of repair. A specific small molecule inhibitor for MPG activity would be a valuable biochemical tool for understanding these repair mechanisms. By screening several small molecule chemical libraries, we identified a natural polyphenolic compound, morin hydrate, which inhibits MPG activity specifically (IC50=2.6µM). Detailed mechanism analysis showed that morin hydrate inhibited substrate DNA binding of MPG, and eventually the enzymatic activity of MPG. Computational docking studies with an x-ray derived MPG structure as well as comparison studies with other structurally-related flavonoids offer a rationale for the inhibitory activity of morin hydrate observed. The results of this study suggest that the morin hydrate could be an effective tool for studying MPG function and it is possible that morin hydrate and its derivatives could be utilized in future studies focused on the role of MPG in human disease.

Allelic variation in KIR2DL3 generates a KIR2DL2-like receptor with increased binding to its HLA-C ligand.

Although extensive homology exists between their extracellular domains, NK cell inhibitory receptors killer Ig-like receptor (KIR) 2DL2*001 and KIR2DL3*001 have previously been shown to differ substantially in their HLA-C binding avidity. To explore the largely uncharacterized impact of allelic diversity, the most common KIR2DL2/3 allelic products in European American and African American populations were evaluated for surface expression and binding affinity to their HLA-C group 1 and 2 ligands. Although no significant differences in the degree of cell membrane localization were detected in a transfected human NKL cell line by flow cytometry, surface plasmon resonance and KIR binding to a panel of HLA allotypes demonstrated that KIR2DL3*005 differed significantly from other KIR2DL3 allelic products in its ability to bind HLA-C. The increased affinity and avidity of KIR2DL3*005 for its ligand was also demonstrated to have a larger impact on the inhibition of IFN-γ production by the human KHYG-1 NK cell line compared with KIR2DL3*001, a low-affinity allelic product. Site-directed mutagenesis established that the combination of arginine at residue 11 and glutamic acid at residue 35 in KIR2DL3*005 were critical to the observed phenotype. Although these residues are distal to the KIR/HLA-C interface, molecular modeling suggests that alteration in the interdomain hinge angle of KIR2DL3*005 toward that found in KIR2DL2*001, another strong receptor of the KIR2DL2/3 family, may be the cause of this increased affinity. The regain of inhibitory capacity by KIR2DL3*005 suggests that the rapidly evolving KIR locus may be responding to relatively recent selective pressures placed upon certain human populations.

Novel Methodology for the Design of Personalized Cancer Vaccine Targeting Neoantigens: Application to Pancreatic Ductal Adenocarcinoma.

Personalized cancer vaccines have emerged as a promising avenue for cancer treatment or prevention strategies. This approach targets the specific genetic alterations in individual patient's tumors, offering a more personalized and effective treatment option. Previous studies have shown that generalized peptide vaccines targeting a limited scope of gene mutations were ineffective, emphasizing the need for personalized approaches. While studies have explored personalized mRNA vaccines, personalized peptide vaccines have not yet been studied in this context. Pancreatic ductal adenocarcinoma (PDAC) remains challenging in oncology, necessitating innovative therapeutic strategies. In this study, we developed a personalized peptide vaccine design methodology, employing RNA sequencing (RNAseq) to identify prevalent gene mutations underlying PDAC development in a patient solid tumor tissue. We performed RNAseq analysis for trimming adapters, read alignment, and somatic variant calling. We also developed a Python program called SCGeneID, which validates the alignment of the RNAseq analysis. The Python program is freely available to download. Using chromosome number and locus data, SCGeneID identifies the target gene along the UCSC hg38 reference set. Based on the gene mutation data, we developed a personalized PDAC cancer vaccine that targeted 100 highly prevalent gene mutations in two patients. We predicted peptide-MHC binding affinity, immunogenicity, antigenicity, allergenicity, and toxicity for each epitope. Then, we selected the top 50 and 100 epitopes based on our previously published vaccine design methodology. Finally, we generated pMHC-TCR 3D molecular model complex structures, which are freely available to download. The designed personalized cancer vaccine contains epitopes commonly found in PDAC solid tumor tissue. Our personalized vaccine was composed of neoantigens, allowing for a more precise and targeted immune response against cancer cells. Additionally, we identified mutated genes, which were also found in the reference study, where we obtained the sequencing data, thus validating our vaccine design methodology. This is the first study designing a personalized peptide cancer vaccine targeting neoantigens using human patient data to identify gene mutations associated with the specific tumor of interest.

A Machine-Learning-Driven Pathophysiology-Based New Approach Method for the Dose-Dependent Assessment of Hazardous Chemical Mixtures and Experimental Validations.

Environmental chemicals, such as PFAS, exist as mixtures and are frequently encountered at varying concentrations, which can lead to serious health effects, such as cancer. Therefore, understanding the dose-dependent toxicity of chemical mixtures is essential for health risk assessment. However, comprehensive methods to assess toxicity and identify the mechanisms of these harmful mixtures are currently absent. In this study, the dose-dependent toxicity assessments of chemical mixtures are performed in three methodologically distinct phases. In the first phase, we evaluated our machine-learning method (AI-HNN) and pathophysiology method (CPTM) for predicting toxicity. In the second phase, we integrated AI-HNN and CPTM to establish a comprehensive new approach method (NAM) framework called AI-CPTM that is targeted at refining prediction accuracy and providing a comprehensive understanding of toxicity mechanisms. The third phase involved experimental validations of the AI-CPTM predictions. Initially, we developed binary, multiclass classification, and regression models to predict binary, categorical toxicity, and toxic potencies using nearly a thousand experimental mixtures. This empirical dataset was expanded with assumption-based virtual mixtures, compensating for the lack of experimental data and broadening the scope of the dataset. For comparison, we also developed machine-learning models based on RF, Bagging, AdaBoost, SVR, GB, KR, DT, KN, and Consensus methods. The AI-HNN achieved overall accuracies of over 80%, with the AUC exceeding 90%. In the final phase, we demonstrated the superior performance and predictive capability of AI-CPTM, including for PFAS mixtures and their interaction effects, through rigorous literature and statistical validations, along with experimental dose-response zebrafish-embryo toxicity assays. Overall, the AI-CPTM approach significantly improves upon the limitations of standalone AI models, showing extensive enhancements in identifying toxic chemicals and mixtures and their mechanisms. This study is the first to develop a hybrid NAM that integrates AI with a pathophysiology method to comprehensively predict chemical-mixture toxicity, carcinogenicity, and mechanisms.

AutoEpiCollect, a Novel Machine Learning-Based GUI Software for Vaccine Design: Application to Pan-Cancer Vaccine Design Targeting PIK3CA Neoantigens.

Previous epitope-based cancer vaccines have focused on analyzing a limited number of mutated epitopes and clinical variables preliminarily to experimental trials. As a result, relatively few positive clinical outcomes have been observed in epitope-based cancer vaccines. Further efforts are required to diversify the selection of mutated epitopes tailored to cancers with different genetic signatures. To address this, we developed the first version of AutoEpiCollect, a user-friendly GUI software, capable of generating safe and immunogenic epitopes from missense mutations in any oncogene of interest. This software incorporates a novel, machine learning-driven epitope ranking method, leveraging a probabilistic logistic regression model that is trained on experimental T-cell assay data. Users can freely download AutoEpiCollectGUI with its user guide for installing and running the software on GitHub. We used AutoEpiCollect to design a pan-cancer vaccine targeting missense mutations found in the proto-oncogene PIK3CA, which encodes the p110ɑ catalytic subunit of the PI3K kinase protein. We selected PIK3CA as our gene target due to its widespread prevalence as an oncokinase across various cancer types and its lack of presence as a gene target in clinical trials. After entering 49 distinct point mutations into AutoEpiCollect, we acquired 361 MHC Class I epitope/HLA pairs and 219 MHC Class II epitope/HLA pairs. From the 49 input point mutations, we identified MHC Class I epitopes targeting 34 of these mutations and MHC Class II epitopes targeting 11 mutations. Furthermore, to assess the potential impact of our pan-cancer vaccine, we employed PCOptim and PCOptim-CD to streamline our epitope list and attain optimized vaccine population coverage. We achieved a world population coverage of 98.09% for MHC Class I data and 81.81% for MHC Class II data. We used three of our predicted immunogenic epitopes to further construct 3D models of peptide-HLA and peptide-HLA-TCR complexes to analyze the epitope binding potential and TCR interactions. Future studies could aim to validate AutoEpiCollect's vaccine design in murine models affected by PIK3CA-mutated or other mutated tumor cells located in various tissue types. AutoEpiCollect streamlines the preclinical vaccine development process, saving time for thorough testing of vaccinations in experimental trials.

AutoPepVax, a Novel Machine-Learning-Based Program for Vaccine Design: Application to a Pan-Cancer Vaccine Targeting EGFR Missense Mutations.

The current epitope selection methods for peptide vaccines often rely on epitope binding affinity predictions, prompting the need for the development of more sophisticated in silico methods to determine immunologically relevant epitopes. Here, we developed AutoPepVax to expedite and improve the in silico epitope selection for peptide vaccine design. AutoPepVax is a novel program that automatically identifies non-toxic and non-allergenic epitopes capable of inducing tumor-infiltrating lymphocytes by considering various epitope characteristics. AutoPepVax employs random forest classification and linear regression machine-learning-based models, which are trained with datasets derived from tumor samples. AutoPepVax, along with documentation on how to run the program, is freely available on GitHub. We used AutoPepVax to design a pan-cancer peptide vaccine targeting epidermal growth factor receptor (EGFR) missense mutations commonly found in lung adenocarcinoma (LUAD), colorectal adenocarcinoma (CRAD), glioblastoma multiforme (GBM), and head and neck squamous cell carcinoma (HNSCC). These mutations have been previously targeted in clinical trials for EGFR-specific peptide vaccines in GBM and LUAD, and they show promise but lack demonstrated clinical efficacy. Using AutoPepVax, our analysis of 96 EGFR mutations identified 368 potential MHC-I-restricted epitope-HLA pairs from 49,113 candidates and 430 potential MHC-II-restricted pairs from 168,669 candidates. Notably, 19 mutations presented viable epitopes for MHC I and II restrictions. To evaluate the potential impact of a pan-cancer vaccine composed of these epitopes, we used our program, PCOptim, to curate a minimal list of epitopes with optimal population coverage. The world population coverage of our list ranged from 81.8% to 98.5% for MHC Class II and Class I epitopes, respectively. From our list of epitopes, we constructed 3D epitope-MHC models for six MHC-I-restricted and four MHC-II-restricted epitopes, demonstrating their epitope binding potential and interaction with T-cell receptors. AutoPepVax's comprehensive approach to in silico epitope selection addresses vaccine safety, efficacy, and broad applicability. Future studies aim to validate the AutoPepVax-designed vaccines with murine tumor models that harbor the studied mutations.

Tyrosine-phosphorylated caveolin-1 (Tyr-14) increases sensitivity to paclitaxel by inhibiting BCL2 and BCLxL proteins via c-Jun N-terminal kinase (JNK).

Paclitaxel, an anti-microtubule agent, is an effective chemotherapeutic drug in breast cancer. Nonetheless, resistance to paclitaxel remains a major clinical challenge. The need to better understand the resistant phenotype and to find biomarkers that could predict tumor response to paclitaxel is evident. In estrogen receptor α-positive (ER(+)) breast cancer cells, phosphorylation of caveolin-1 (CAV1) on Tyr-14 facilitates mitochondrial apoptosis by increasing BCL2 phosphorylation in response to low dose paclitaxel (10 nM). However, two variants of CAV1 exist: the full-length form, CAV1α (wild-type CAV1 or wtCAV1), and a truncated form, CAV1ß. Only wtCAV1 has the Tyr-14 region at the N terminus. The precise cellular functions of CAV1 variants are unknown. We now show that CAV1 variants play distinct roles in paclitaxel-mediated cell death/survival. CAV1ß expression is increased in paclitaxel-resistant cells when compared with sensitive cells. Expression of CAV1ß in sensitive cells significantly reduces their responsiveness to paclitaxel. These activities reflect an essential role for Tyr-14 phosphorylation because wtCAV1 expression, but not a phosphorylation-deficient mutant (Y14F), inactivates BCL2 and BCLxL through activation of c-Jun N-terminal kinase (JNK). MCF-7 cells that express Y14F are resistant to paclitaxel and are resensitized by co-treatment with ABT-737, a BH3-mimetic small molecule inhibitor. Using structural homology modeling, we propose that phosphorylation on Tyr-14 enables a favorable conformation for proteins to bind to the CAV1 scaffolding domain. Thus, we highlight novel roles for CAV1 variants in cell death; wtCAV1 promotes cell death, whereas CAV1ß promotes cell survival by preventing inactivation of BCL2 and BCLxL via JNK in paclitaxel-mediated apoptosis.

Mechanisms of unphosphorylated STAT3 transcription factor binding to DNA.

Phosphorylation of signal transducer and activator of transcription 3 (STAT3) on a single tyrosine residue in response to growth factors, cytokines, interferons, and oncogenes activates its dimerization, translocation to the nucleus, binding to the interferon γ (gamma)-activated sequence (GAS) DNA-binding site and activation of transcription of target genes. STAT3 is constitutively phosphorylated in various cancers and drives gene expression from GAS-containing promoters to promote tumorigenesis. Recently, roles for unphosphorylated STAT3 (U-STAT3) have been described in response to cytokine stimulation, in cancers, and in maintenance of heterochromatin stability. However, the mechanisms underlying U-STAT3 binding to DNA has not been fully investigated. Here, we explore STAT3-DNA interactions by atomic force microscopy (AFM) imaging. We observed that U-STAT3 molecules bind to the GAS DNA-binding site as dimers and monomers. In addition, we observed that U-STAT3 binds to AT-rich DNA sequence sites and recognizes specific DNA structures, such as 4-way junctions and DNA nodes, within negatively supercoiled plasmid DNA. These structures are important for chromatin organization and our data suggest a role for U-STAT3 as a chromatin/genome organizer. Unexpectedly, we found that a C-terminal truncated 67.5-kDa STAT3 isoform recognizes single-stranded spacers within cruciform structures that also have a role in chromatin organization and gene expression. This isoform appears to be abundant in the nuclei of cancer cells and, therefore, may have a role in regulation of gene expression. Taken together, our data highlight novel mechanisms by which U-STAT3 binds to DNA and supports U-STAT3 function as a transcriptional activator and a chromatin/genomic organizer.

Predicting Dose-Dependent Carcinogenicity of Chemical Mixtures Using a Novel Hybrid Neural Network Framework and Mathematical Approach.

This study addresses the challenge of assessing the carcinogenic potential of hazardous chemical mixtures, such as per- and polyfluorinated substances (PFASs), which are known to contribute significantly to cancer development. Here, we propose a novel framework called HNNMixCancer that utilizes a hybrid neural network (HNN) integrated into a machine-learning framework. This framework incorporates a mathematical model to simulate chemical mixtures, enabling the creation of classification models for binary (carcinogenic or noncarcinogenic) and multiclass classification (categorical carcinogenicity) and regression (carcinogenic potency). Through extensive experimentation, we demonstrate that our HNN model outperforms other methodologies, including random forest, bootstrap aggregating, adaptive boosting, support vector regressor, gradient boosting, kernel ridge, decision tree with AdaBoost, and KNeighbors, achieving a superior accuracy of 92.7% in binary classification. To address the limited availability of experimental data and enrich the training data, we generate an assumption-based virtual library of chemical mixtures using a known carcinogenic and noncarcinogenic single chemical for all the classification models. Remarkably, in this case, all methods achieve accuracies exceeding 98% for binary classification. In external validation tests, our HNN method achieves the highest accuracy of 80.5%. Furthermore, in multiclass classification, the HNN demonstrates an overall accuracy of 96.3%, outperforming RF, Bagging, and AdaBoost, which achieved 91.4%, 91.7%, and 80.2%, respectively. In regression models, HNN, RF, SVR, GB, KR, DT with AdaBoost, and KN achieved average R2 values of 0.96, 0.90, 0.77, 0.94, 0.96, 0.96, and 0.97, respectively, showcasing their effectiveness in predicting the concentration at which a chemical mixture becomes carcinogenic. Our method exhibits exceptional predictive power in prioritizing carcinogenic chemical mixtures, even when relying on assumption-based mixtures. This capability is particularly valuable for toxicology studies that lack experimental data on the carcinogenicity and toxicity of chemical mixtures. To our knowledge, this study introduces the first method for predicting the carcinogenic potential of chemical mixtures. The HNNMixCancer framework offers a novel alternative for dose-dependent carcinogen prediction. Ongoing efforts involve implementing the HNN method to predict mixture toxicity and expanding the application of HNNMixCancer to include multiple mixtures such as PFAS mixtures and co-occurring chemicals.