Back

A Meta-model for ADMET Property Prediction Analysis

Padi, S.; Cardone, A.; Sriram, R. D.

2023-12-07 bioinformatics
10.1101/2023.12.05.570279 bioRxiv
Show abstract

In drug discovery analysis chemical absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties play a critical role. These properties allow the quantitative evaluation of a designed drugs efficacy. Several machine learning models have been designed for the prediction of ADMET properties. However, no single method seems to enable the accurate prediction of these properties. In this paper, we build a meta-model that learns the best possible way to combine the scores from multiple heterogeneous machine learning models to effectively predict the ADMET properties. We evaluate the performance of our proposed model against the Therapeutics Data Commons (TDC) ADMET benchmark dataset. The proposed meta-model outperforms state-of-the-art methods such as XGBoost in the TDC leaderboard, and it ranks first in five and in the top three positions for fifteen out of twenty-two prediction tasks.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.