Auto-omics for healthcare and drug discovery

An automated explainable bioinformatics and AI workflow for multi-omic, clinical and experimental data, applied to healthcare and drug discovery problems.


We’re developing an explainable AI solution that combines multi-modal data to predict a phenotype (such as a disease’s progression or the response to drugs or treatment) and to infer insightful biomarkers. We process data types ranging from multi-omic data (such as metagenomic and metatranscriptomic, metabolomic, genomic, and transcriptomic data), clinical data, demographic data and medicinal data to pharmacology data (like functional experimental assays). Inferred predictive biomarkers can be in the form of key contributing microbiome-markers , including microbial species or functions, genomic markers like SNPs, transcriptomic markers (like gene expression counts), metabolomic, integrated multi-modal markers (including species-function, species-metabolite, or species-function-metabolite), or clinical markers that drive accurate phenotype predictions. Our core methods can be applied to investigate a wide range of phenotypes, but here our primary focus is on a series of multi-modal use-cases (which include omics data) that determine patient-specific drug responses, disease sub-types, or progression or to inform clinical trial precision medicine strategies. Our goal is to determine whether we can use multi-modal data to build biomarker profiles linked to patient treatment response or disease sub-types.

Some of our current projects:

  • Developing an explainable ML workflow to understand patient-specific responses to drug treatment for inflammatory bowel disease and identifying the treatments most suitable for clinical trials, and the patients most likely to respond to treatment.

  • Developing a novel solution combining unsupervised and supervised learning approaches to investigate disease sub-types or stages.

  • Developing an end-to-end solution that identifies key biomarkers of a patient's menopausal stage from analysis of the microbiome, using bioinformatics and ML methods.

  • Building an AI-enabled solution that integrates multi-omics and demographic data to identify multi-omic biomarkers of skin phenotypes, allowing clients to develop personalized skincare solutions.

journal.pone.0263248.g001.PNGSchematic of example multi-omic integration workflow for drug testing-relating to project 1 above. From microbiome AI workflow-relating to project 4 above. From




Related projects

Auto-omics for climate and sustainability

An automated explainable bioinformatics and AI workflow for multi-omic, climate and environmental data, applied to sustainability problems e.g., nature-based carbon capture.
Host-pathogen interactions.png

Host-pathogen interactions for healthcare and drug discovery

An automated explainable bioinformatics and deep learning workflow to elucidate microbiome-metabolite relationships during a host-pathogen interaction.