This study addresses the challenge of assessing the carcinogenic potential of hazardous chemical mixtures, such as PFAS, 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 and enables the creation of classification models for binary and multiclass classification and regression. 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 non-carcinogenic 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 achieve 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 multi-mixtures.

Environmental chemicals, including PFAS (per- and polyfluoroalkyl substances), pesticides, industrial chemicals, and consumer products, commonly exist as mixtures. These substances are frequently exposed or co-exposed in varying concentrations, leading to potentially hazardous health effects such as cancer in humans. Thus, understanding the dose-dependent toxicity of chemical mixtures is important for assessing health risks. In this context, comprehensive methods for assessing the toxicity and identifying the mechanisms of harmful chemical mixtures are currently lacking. Here, 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, 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 virtual mixtures compensating the lack of experimental data and broadening the scope of the dataset. For comparison, we also developed additional machine learning models based on RF, Bagging, AdaBoost, SVR, GB, KR, DT, KN, and Consensus methods. The models achieved overall accuracies of over 80% with AUC values exceeding 90%. The regression models achieved an R2 >0.88. In the second phase, we innovatively integrated HNN-derived toxicity predictions with Z-scores from CPTM, resulting method called AI-CPTM. In the final phase, we demonstrated the superior performance of AI-CPTM through rigorous literature and statistical performance validations. Additionally, the predictive capability of AI-CPTM, including for PFAS mixtures and their interaction effects, was demonstrated by experimental validations using dose-response zebrafish embryo toxicity assays. Overall, the AI-CPTM approach significantly improves upon the limitations of standalone models and has shown extensive enrichments in the identification of toxic chemicals and mixtures. Further experimental studies involving human cell models, patient-derived xenografts, and investigations into the toxicity of multiple mixtures are currently underway.