Big Data Approaches to Statistically Characterizing Biochemistries in the Context of their Planetary Environment

Abstract:

The existence of curated databases cataloging biochemical diversity across Earth is enabling new directions in biosignature science. In particular, whereas biosignatures have traditionally been relegated to focusing on specific molecules universal to biochemistry, we are now in a position to analyze statistical patterns in the the use of elements, molecular properties, reaction types, and of the network structure of pathways and metabolisms across individuals and ecosystems to derive new agnostic biosignatures. In this talk, we present results from adopting a multi-variate, big data approach regularizing distributions in elemental use, molecular properties, reactions and reaction networks across multiple scales of biochemical diversity sampled from across Earth, including data from > 32,000 genomes and metagenomes from the Joint Genome Institute (JGI), annotate using the Kyoto Encyclopedia of Genes and Genomes (KEGG). We also discuss applications to understanding the biogeochemical evolution of life on Earth when statistical patterns are correlated with environmental metadata, and to the development new statistically robust biosignature frameworks.