Deep learning algorithm integrates multiomics data to identify unique functional molecules and diagnostic biomarkers from the human microbiome.

We live in a deep symbiosis with trillions of microbes, together with their genes and secreted molecules it is referred to as the human microbiome. Mounting evidence suggests the crucial role of the microbiome in shaping human health or mediating diseases although the mechanistic underpinning is lacking. Discovery of functional genes or molecules from the microbiome holds a promise to develop unique therapeutics and diagnostic biomarkers for multiple diseases from metabolic disorders to inflammatory autoimmune diseases and malignancies. In this research we present a novel deep learning algorithm that integrates metabolomic and transcriptomic data to annotate the microbiome function ultimately unlocking its chemistry. The algorithm pinpoints functional molecules directly in the mass spectrometry data enabling downstream isolation, structural elucidation, and in-depth biological evaluation. Guided by the new tool we identified and fully characterised a novel immune modulatory molecule, we named it limousine, encoded in a rare poorly understood human-associated microbe. Our data shows that this molecule is consistently correlated with the onset and progression of inflammatory bowel syndrome and could be a potential microbiome- based biomarker. We envision our tool to be a driving force to identify active molecules in silico and guide the downstream structural characterisation. The algorithm is applicable to all microbiome systems with potential application in all fields from agriculture to medicine.

 

MissingLinkText

Conflict of Interest

The authors declare no conflict of interest.

Publikationsverlauf

Artikel online veröffentlicht:
16. November 2023