Improving personalised disease diagnostics

By analysing microbiome data, IBM Research scientists based at the STFC Hartree Centre are working to predict whether an individual has an illness, a predisposition to a disease, or how a disease has developed in a non-invasive way.

Challenge

The applications of metagenomics (the study of genomic diversity of microbes) vary from food safety to environmental studies. In healthcare, characterising composition and activity of metagenomes in individuals (both healthy and disease subjects) is important to understand the potential role of the microbiome in disease progression. The microbiome is a collection of genetic material and interactions of a community of microorganisms (bacteria, fungi, viruses) inhabiting a particular environment such as the human gut. We need a non-invasive way of investigating how microbes in the microbiome may affect our health.​

Approach

The sequencing of a specific marker gene, 16S rRNA, allows us to classify microbes in to Operational Taxonomy Units (OTUs), each unit representing a different species. This high dimensional metagenomics data, related to the relative abundance of the microbial species, are represented in sparse OTUs tables. We then apply normalization techniques, such as RoDEO (Robust Differential Gene Expression), to the OTU tables. RoDEO, is a software originally developed to compute differential expressed genes and when applied in metagenomics it allows to identify bacterial species enabling us to distinguish between different phenotypic groups. This allows us to identify bacterial species that may influence disease progression. By applying RoDEO normalisation and feature selection combined with machine learning methods, we have been able to accurately predict phenotypes or traits of host organisms. For example, we have been able to predict if individuals had Type 2 Diabetes or Normal Glucose Tolerance.

Benefits

These data science techniques help to better understand what the human gut microbiome can tell us about our health, enabling us to predict an individual’s susceptibility to disease and helping medical professionals to reach a disease diagnosis. Making predictions of whether someone is healthy, has a condition or a predisposition to a condition from their gut microbiome would help diagnostics and provide valuable insight for the design of personalised treatments.

“Although we can’t see microbes, trillions of them live inside our bodies where they can play a positive or negative role in determining our well-being. An imbalance in the microbial community can be linked to development of metabolic, respiratory and gastrointestinal diseases. Understanding these interactions can help with designing effective, targeted treatments focused on re-equilibrating the microbiome.”

Dr. Anna Paola Carrieri, IBM Research​