You are here:
Feasibility Analysis of Incorporating In-Vitro Toxicokinetic Data as a Surrogate for In-Vivo Data for Read-across Predictions (ASCCT meeting)
Citation:
Pradeep, P., R. Judson, AND G. Patlewicz. Feasibility Analysis of Incorporating In-Vitro Toxicokinetic Data as a Surrogate for In-Vivo Data for Read-across Predictions (ASCCT meeting). Presented at ASCCT, RTP, NC, September 29 - 30, 2016.
Impact/Purpose:
poster presented at ASCCT meeting in RTP, NC.
Description:
The underlying principle of read-across is that biological activity is a function of physical and structural properties of chemicals. Analogs are typically identified on the basis of structural similarity and subsequently evaluated for their use in read-across on the basis of their bioavailability, reactivity and metabolic similarity. While the concept of similarity is the major tenet in grouping chemicals for read-across, a critical consideration is to evaluate if structural differences significantly impact toxicological activity. This is a key source of uncertainty in read-across predictions. We hypothesize that inclusion of toxicokinetic (TK) information will reduce the uncertainty in read-across predictions. TK information can help substantiate whether chemicals within a category have similar ADME properties and, hence, increase the likelihood of exhibiting similar toxicological properties. This current case study is part of a larger study aimed at performing a systematic assessment of the extent to which in-vitro TK data can obviate in-vivo TK data, while maintaining or increasing scientific confidence in read-across predictions. The analysis relied on a dataset of ~7k chemicals with predicted exposure data (chemical inventory), of which 819 chemicals had rat and/or human in-vitro TK data (analog inventory), and 33 chemicals had rat in-vivo TK data (target inventory). The set of chemicals with human in vitro TK data was investigated to determine whether structurally related chemicals had similar intrinsic clearance. An unsupervised feature selection was performed (using Chemotyper and PubChem fingerprints) on chemicals in the analog inventory to remove the features with low variance. Unsupervised clustering was performed on the reduced feature set using the K-means algorithm. Preliminary results (using both fingerprints) show a correlation between structural fingerprint-based clusters and intrinsic clearance. The chemical inventory was then explored using principal component analysis based on both fingerprints to help select target chemicals that were representative of the inventory, and also associated with in-vivo TK data. 11 target chemicals were found to have at least 1 analog with in-vitro data. Next steps will include an analysis of the correspondence between in-vitro and in-vivo TK data using the target chemicals identified.
URLs/Downloads:
HTTK_KFC_RSJ_FINAL.PDF (PDF, NA pp, 787.647 KB, about PDF)ABSTRACT_ASCCT2016__FOR CLEARANCE_250816 KMC_REVISED POSTKMC.PDF (PDF, NA pp, 37.184 KB, about PDF)