Anticipating the chemistry of Life

ChallengesAdvancements WP1 -The first version of the Markov Random Field works. -Inference is fast. With a test set of 17’000 molecules and 4000 species, each MCMC iteration performs 7×10⁷ updates in 15 seconds. Side project -Currently working on the in silico fragmentation of the 3M compounds. -Building a GUI for the exploration of the expanded dataset (e.g. taxon sources, reactions, spectra) 0 0 1 1 1 0 0 1 0 0 1 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 1 0 0 0 1 1 1 0 0 1 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 P( | ) 0 0 1 1 1 0 0 1 0 0 1 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 1 0 Extraction process Liquid Chromatography Mass Spectrometry Anticipating the chemistry of Life Markov Random Field -When inferring only with LOTUS data, every prediction goes to zero. This is probably due to the extreme sparsity of LOTUS, leading the current model to consider all data as noise. Mass spec data -The overall process ranging from sample extraction to MS data is highly dimensional and complex. Information of data loss at each steps is not easily rationalized. Ongoing work Marco Visani1,2, Madleina Caduff1,2, Daniel Wegmann1,2, Pierre-Marie Allard1,2. 1 Department of Biology, University of Fribourg, Fribourg, Switzerland. 2 Swiss Institute of Bioinformatics, Lausanne, Switzerland. Side-project In silico"metabolic expansion of a natural products database y y LOTUS Anticipated LOTUS biological organisms molecular structures < < NMR Structural determination WP1 Model NPs occurences WP2 Model MS analysis Untargeted metabolomics [Previous Knowledge] Academic corpus of publication on Natural Products Chemistry [Novel Knowledge] UniFr Botanical Garden Targeted isolations Isolation WP4 Validation WP4 Validation Experimental evidences Hot off the press articles WP4 Validation Created by Azam Ishaq from the Noun Project Bioprospection Drug discovery Biodiversity Conservation Where can we find novel antibiotic structures ? Molecular arguments for conservation : why should we prioritize the conservation of this biome ? www.dbgi.org DBGI The Digital Botanical Gardens Initiative Application example Application example WP3 Predicting occurences across the tree of life www.earthmetabolome.org I N I T I AT I V E EARTH METABOLOME Public metabolomics datasets ViMMS molecular structures biological organisms General framework LOTUS From the molecules in LOTUS and a set of enzymatic reaction rules, we generated in silico 3 million plausible compounds. 16’991 of which are real natural products found in other databases.