Gene Target Prediction of Environmental Chemicals Using Coupled Matrix-Matrix Completion.

Human exposure to toxic chemicals presents a huge health burden. Key to understanding chemical toxicity is knowledge of the molecular target(s) of the chemicals. Because a comprehensive safety assessment for all chemicals is infeasible due to limited resources, a robust computational method for discovering targets of environmental exposures is a promising direction for public health research. In this study, we implemented a novel matrix completion algorithm named coupled matrix-matrix completion (CMMC) for predicting direct and indirect exposome-target interactions, which exploits the vast amount of accumulated data regarding chemical exposures and their molecular targets. Our approach achieved an AUC of 0.89 on a benchmark data set generated using data from the Comparative Toxicogenomics Database. Our case studies with bisphenol A and its analogues, PFAS, dioxins, PCBs, and VOCs show that CMMC can be used to accurately predict molecular targets of novel chemicals without any prior bioactivity knowledge. Our results demonstrate the feasibility and promise of computationally predicting environmental chemical-target interactions to efficiently prioritize chemicals in hazard identification and risk assessment.

Whole-genome CpG-resolution DNA Methylation Profiling of HNSCC Reveals Distinct Mechanisms of Carcinogenesis for Fine-scale HPV+ Cancer Subtypes.

DNA methylation is a vital early step in carcinogenesis. Most findings of aberrant DNA methylation in head and neck squamous cell carcinomas (HNSCC) are array based with limited coverage and resolution, and mainly explored by human papillomavirus (HPV) status, ignoring the high heterogeneity of this disease. In this study, we performed whole-genome bisulfite sequencing on a well-studied HNSCC cohort (n = 36) and investigated the methylation changes between fine-scaled HNSCC subtypes in relation to genomic instability, repetitive elements, gene expression, and key carcinogenic pathways. The previously observed hypermethylation phenotype in HPV-positive (HPV+) tumors compared with HPV-negative tumors was robustly present in the immune-strong (IMU) HPV+ subtype but absent in the highly keratinized (KRT) HPV+ subtype. Methylation levels of IMU tumors were significantly higher in repetitive elements, and methylation showed a significant correlation with genomic stability, consistent with the IMU subtype having more genomic stability and better prognosis. Expression quantitative trait methylation (cis-eQTM) analysis revealed extensive functionally-relevant differences, and differential methylation pathway analysis recapitulated gene expression pathway differences between subtypes. Consistent with their characteristics, KRT and HPV-negative tumors had high regulatory potential for multiple regulators of keratinocyte differentiation, which positively correlated with an expression-based keratinization score. Together, our findings revealed distinct mechanisms of carcinogenesis between subtypes in HPV+ HNSCC and uncovered previously ignored epigenomic differences and clinical implications, illustrating the importance of fine-scale subtype analysis in cancer. Significance: This study revealed that the previously observed hypermethylation of HPV(+) HNSCC is due solely to the IMU subtype, illustrating the importance of fine-scale subtype analysis in such a heterogeneous disease. Particularly, IMU has significantly higher methylation of transposable elements, which can be tested as a prognosis biomarker in future translational studies.

Protocol for using Ciclops to build models trained on cross-platform transcriptome data for clinical outcome prediction.

Designing robust, generalizable models based on cross-platform data to predict clinical outcomes remains challenging. Building explainable models is important because models may perform differently depending on the conditions of the samples. Here, we describe the use of Ciclops (cross-platform training in clinical outcome predictions), freely available software that can build explainable models to deliver across cross-platform datasets for predicting clinical outcomes. This protocol also utilizes SHAP, a post-training analysis allowing for assessing potential biomarkers of the clinical outcome under study. For complete details on the use and execution of this protocol, please refer to Zhang et al. (2022).