Host-pathogen interactions for healthcare and drug discovery
We’re developing a scalable multi-modal deep learning framework to infer the complex and dynamic relationships between the microbiome and metabolome of a host system during a pathogenic infection. Multi-omics is an important paradigm to understand a biological system from a combined perspective derived from a range of diverse omics evidence. Within the context of vaccine development for a pathogen induced disease, it is playing an increasingly important role in understanding host-pathogen interaction and efficacy of potential vaccine candidates.
Understanding the microbiome-metabolite relationship is a crucial element in foundation models for healthcare as the microbiome plays a critical role in influencing human health and disease, where alterations in the microbiome-metabolite relationship have been linked to several diseases, including obesity, diabetes, and inflammatory bowel disease. To drive progress in the rapidly advancing fields of personalized medicine and precision nutrition, it’s essential to build a cohesive and explainable view on this complex yet vital biological relationship.
We explore the relationships between microbiome and metabolite interaction using advanced, data-driven AI techniques. The inference of microbiome-metabolite relationship is an open area of study, and is emerging as a regular theme in understanding healthcare due to the reduced cost of data generation, as well as the effectiveness of metabolites as a proxy for easily measurable phenotypes. We bring together bioinformatics, network biology, and machine learning to develop algorithms and tools for data processing, inference, and prediction steps that can capture the host-pathogen dynamics and help infer the microbiome-metabolome relationship. We plan to introduce model explainability from both biological and computational perspectives. Our future aim is to use the computational predictions to develop further in-vivo experiments to check for the interpretability of models, and validate the results through follow-on experimental evidence.
We’re using time-course, tissue specific microbiome and metabolite data collected from experiments designed to understand host-pathogen interactions. Our team is developing a multi-modal deep learning framework to infer the complex and dynamical relationships between the microbiome and metabolome of a host system during a pathogenic infection.
- Using human in vitro transcriptome analysis to build trustworthy machine learning models for prediction of animal drug toxicity
- Combining human cell line transcriptome analysis and Bayesian inference to build trustworthy machine learning models for prediction of animal toxicity in drug development