AI-assisted ADMET Property Prediction Services

ADMET stands as an abbreviation for absorption, distribution, metabolism, excretion, and toxicity in pharmacokinetics and pharmacology. These properties describe the reaction of a drug medication within an organism, human, or animal, which helps scientists evaluate their drug candidates and select compounds that are likely to have the desired effects on patients. Building ADMET profiles in a wet-lab setting are very costly and time-consuming. The growing performance of machine learning algorithms and the increased availability of the ADMET dataset makes it possible to predict the compound’s properties in silico. Based on the advanced AI-assisted platform, Creative Biolabs can help our customers screen thousands of compounds and make it possible to screen compounds at any stage of the drug discovery process.

ADMET structure−activity relationship database.Fig.1 ADMET structure−activity relationship database. (Cheng, et al., 2012)

AI-assisted Aqueous Solubility Prediction

As one of the most fundamental properties of drug-like molecules, aqueous solubility has been predicted by several in silico models. Using a combination of descriptors and statistical approaches, some group developed quantitative structure-activity relationship (QSAR) models for solubility prediction based on FDAMDD and PHYSPROP databases. In addition, using a support vector machine (SVM) algorithm with reduction and recombination feature selection methods, some researchers built binary classification models of aqueous solubility. They used the largest known public compounds with experimental solubility data and the overall accuracy of the best model was 84%.

AI-assisted Blood-Brain Barrier Penetration Prediction

During the process of drug discovery, it is important to determine whether a drug will penetrate and distribute within the central nervous system (CNS) with the requisite pharmacokinetic and pharmacodynamic performance. For evaluating CNS penetration, a variety of in vivo and in vitro methods have been developed and applied to advance drug candidates with the desired properties. However, these experimental approaches are typically expensive and time-consuming. To address virtual screening and prospective design, in silico methods to predict CNS penetration from chemical structures have been developed.

AI-assisted Cytochrome P450 Inhibition Prediction

Patients often take multiple medications at once, and if the medications compete with one another for the same enzymes to be metabolized, it could have catastrophic consequences. Therefore, an early understanding of the potential CYP interaction of drug candidates is crucial and new chemical entities (NCEs) should be investigated for CYP inhibition as early as possible in drug research. In silico methods to predict the potential CYP inhibition of drugs are appealing due to their often-low cost and they may be applied to entire chemical libraries at the outset of the drug discovery process. Moreover, predictions can be made on virtual compounds.

Distribution of the predictions with respect to probability of class membership.Fig.2 Distribution of the predictions with respect to probability of class membership. (Jensen, et al., 2007)

AI-assisted Hepatotoxicity Prediction

Drug-induced liver injury (DILI) is one of the leading causes of drug failure in clinical trials, which seriously impeded the development of new drugs. Early assessment of DILI risk for drug candidates is considered an effective strategy to reduce attrition rates for drug discovery. There have been continuous attempts in the prediction of DILI, but successfully predicting DILI remains a huge challenge. Traditional methods for hepatotoxicity assessment are the experiment, which is time-consuming and labor-intensive. In silico methods for predicting the DILI based on the high-quality QSAR model have been developed and archived satisfactory performance.

AI-assisted Human Intestinal Absorption Prediction

Oral administration is the predominant route for medication and about 56% of unique drugs approved by the FDA in 2018 were orally administrated. Therefore, drug absorption is one of the key factors to be considered in drug discovery and development as well as in practical application. Drug absorption mainly relies on solubility and human intestinal absorption (HIA). Prediction of HIA is a major goal in the development of oral drugs. Due to the extremely fast throughput and low cost, in silico methods especially provide valuable advantages in ADME/Tox profiling and have been seamlessly integrated into drug discovery and development. As such, a in silico model is very useful for the prediction of intestinal permeability.

AI-assisted Plasma Protein Binding Prediction

Generally, it is important to maintain the biostability of a drug within the proper range in drug discovery. Plasma protein binding (PPB) is the most important index of biostability, which is strongly related to the ADMET properties of drugs. The majority of the time, only unbound drug components may enter tissues, which then interact with the target proteins and are finally excreted from the blood. The experimental measurements of PPB are expensive and time-consuming. Therefore, in silico methods have been developed to estimate the PPB values of candidate compounds computationally in the early stages.


  1. Cheng, F.; et al. admetSAR: a comprehensive source and free tool for assessment of chemical ADMET properties. 2012.
  2. Jensen, B.F.; et al. In silico prediction of cytochrome P450 2D6 and 3A4 inhibition using Gaussian kernel weighted k-nearest neighbor and extended connectivity fingerprints, including structural fragment analysis of inhibitors versus noninhibitors. Journal of medicinal chemistry. 2007, 50(3): 501-511.

Let's Get Started

Contact Us



Follow us on

Copyright © 2022 Creative Biolabs. All Rights Reserved.