%0 Journal Article
%T A data-driven integrative platform for computational prediction of toxin biotransformation with a case study.
%A Zhang D
%A Tian Y
%A Tian Y
%A Xing H
%A Liu S
%A Zhang H
%A Ding S
%A Cai P
%A Sun D
%A Zhang T
%A Hong Y
%A Dai H
%A Tu W
%A Chen J
%A Wu A
%A Hu QN
%J J Hazard Mater
%V 408
%N 0
%D 04 2021 15
%M 33360695
%F 14.224
%R 10.1016/j.jhazmat.2020.124810
%X Recently, biogenic toxins have received increasing attention owing to their high contamination levels in feed and food as well as in the environment. However, there is a lack of an integrative platform for seamless linking of data-driven computational methods with 'wet' experimental validations. To this end, we constructed a novel platform that integrates the technical aspects of toxin biotransformation methods. First, a biogenic toxin database termed ToxinDB (http://www.rxnfinder.org/toxindb/), containing multifaceted data on more than 4836 toxins, was built. Next, more than 8000 biotransformation reaction rules were extracted from over 300,000 biochemical reactions extracted from ~580,000 literature reports curated by more than 100 people over the past decade. Based on these reaction rules, a toxin biotransformation prediction model was constructed. Finally, the global chemical space of biogenic toxins was constructed, comprising ~550,000 toxins and putative toxin metabolites, of which 94.7% of the metabolites have not been previously reported. Additionally, we performed a case study to investigate citrinin metabolism in Trichoderma, and a novel metabolite was identified with the assistance of the biotransformation prediction tool of ToxinDB. This unique integrative platform will assist exploration of the 'dark matter' of a toxin's metabolome and promote the discovery of detoxification enzymes.