Traditional in silico Quantitative Structure Activity Relationship (QSAR)-based toxicity prediction is susceptible to several limitations, including limited applicability domain with regards to chemical structures, possible inclusion of uninterpretable descriptors, and often lack in mechanistic interpretation of the model. Linking the abundant contemporary in-vitro high throughput screening (HTS) based toxicity assay data to the in-vivo toxicity endpoint, in line with the adverse outcome pathway, shows promises to overcome these limitations. In this study, we have explored the bioassay based Quantitative Biological Activity Relationship (QBAR) model, with in vitro HTS-based temporal transcriptomic and proteomic data of known stress response pathways as the biological descriptors, to predict in vivo toxicity. The study compares the predictive efficacy of the physical-chemical property-based QSAR model, bioassay-based QBAR model, and combined physical chemical property and bioassay-based Quantitative Structure and Biological Activity Relationship (QSBAR) model. Two separate case studies are conducted including cytotoxicity (E. coli EC50) prediction from E. coli-based transcriptomic assay, and genotoxicity (in vivo carcinogenicity) prediction from yeast-based proteomic assay. For the cytotoxicity prediction, the bioassay assisted QBAR models perform better (higher R2) compared to the conventional physical-chemical based QSAR model or the QSBAR models with both physical-chemical and biological descriptors. Both QBAR and QSBAR models are constructed with biodescriptors derived and quantified at three biologically relevant levels — individual biomarker, specific pathway, and cellular level. Among the three levels, the QBAR model at the pathway level yields the best prediction with the highest R2 (0.78), followed by the biomarker level model (R2=0.70). QBAR models also outperform the QSAR, and QSBAR models in predicting the in vivo carcinogenicity. The pathway level QBAR model produces the strongest model (accuracy=75%, sensitivity=77%, and specificity=71%) for the in vivo carcinogenicity prediction. The current study suggests that, bioassay-based QBAR model improves the predictive power of the in silico toxicity prediction and facilitates effective chemical hazard identification and risk monitoring.

In vitro toxicogenomics-derived biodescriptors in in silico QSAR improve the phenotypic toxicity prediction accuracy

Macromolecular Chemistry: The Second Century

[Author Benefits](https://scimeetings.acs.org/?utm_source=pubs_content_marketing&utm_medium=website&utm_campaign=0320_MCF_NPI_Launch_Spring_Homepage&ref=pubs_content_marketing)

[How to use SciMeetings](https://storage.googleapis.com/dl.morressier.com/acs/SciMeetings-ACSSpring2020-HowToGuide_M3020-Final.pdf)

[Frequently Asked Questions](https://storage.googleapis.com/dl.morressier.com/acs/SciMtng_ACSNMSp2020_FAQ_M3120_Final.pdf)

ACS Spring 2020 National Meeting & Expo

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.