• search
ADMET Predictive Models in Aurigene.AI ADMET Predictive Models in Aurigene.AI

ADMET Predictive Models in Aurigene.AI

PUBLISHED ON:
January 30, 2025
CATEGORY:
AI and Drug Discovery

What makes a drug effective and safe for human use?

The answer lies in understanding ADMET, which stands for absorption, distribution, metabolism, excretion, and toxicity. These parameters are critical components of pharmacokinetics, describing how a drug is processed in the body. ADMET properties determine the pharmacokinetic profile of a drug, influencing its efficacy and safety. Optimizing these properties is essential in drug discovery and development, as they affect the drug's absorption into the bloodstream, distribution to tissues, metabolic transformation, excretion from the body, and potential toxicity.

Why is ADMET important?

The primary goal of drug discovery is to identify potential and safer therapeutic compounds for specific diseases. Understanding ADMET is crucial in drug discovery as it helps in predicting the pharmacokinetic and safety profiles of compounds. Thus the ADMET specific data is crucial in reducing the likelihood of nonviable molecules that either do not fall within the acceptable range or are too rigid to be optimized. To enhance this process, researchers increasingly rely on computational methods. Among these, artificial intelligence (AI) and machine learning (ML) have emerged as promising approaches for early screening and optimization, offering significant improvements in efficiency and accuracy.

What is Aurigene.AI?

Aurigene.AI is an advanced platform that leverages AI and machine learning methods to develop highly accurate ADMET prediction models using Industry-Academia trusted datasets to address critical endpoints like lipophilicity, bioavailability, and toxicity. Other features in the platform consists of Multiparameter Optimization (MPO) guided target specific molecule generation; generative-AI enabled de novo molecule design, prediction of binding sites, AI-enabled docking, etc., By offering localized, project-specific solutions and a human-in-the-loop approach, Aurigene.AI accelerates drug discovery with precise, reliable predictions tailored to diverse chemical landscapes.

How Aurigene.AI solves ADMET issues?

ADMET Predictive Models

The ADMET prediction functionality in the Aurigene.AI platform is featured through MPO- ADMET, which predicts ADMET endpoints of a compound. This assists in the prioritization of hits, hit-to-lead, and lead optimization use cases. These ADMET endpoints were developed using data obtained from industry-academia trusted datasets. The models are fine-tuned on various statistical evaluation metrics to benchmark and standardize them against global ADMET models.

The predictive models exhibit high accuracy, with a significant percentage of predictions within 1 log unit of the actual values, demonstrating their precision. Specifically, the models achieved or exceeded a mean average of 95% accuracy rate and an MCC of 0.4, comparable to leading efforts in the field. Additionally, most models surpassed the 0.65 threshold for true positive/true negative rates, indicating robust performance both per class and overall. The confidence metrics at 95% further validate the reliability of these models. For instance, the Lipophilicity model shows a confidence interval of 2.251 to 2.400, and the bioavailability model ranges from 0.796 to 0.887, underscoring the precision of these predictions.

Other models, such as Caco2 and HIA, also demonstrate 100% accuracy within their respective confidence intervals, highlighting their robustness. This high level of accuracy and reliability underscores the models’ effectiveness in predicting ADMET properties, which is essential for identifying viable drug candidates early in the discovery process.

The ADMET predictive models present in the Aurigene.AI platform serve as a crucial filter for cost-effective drug discovery, enabling the strategic inclusion or exclusion of compounds based on their pharmacokinetic and safety profiles. Our human-in-the-loop approach for the DMTA cycle helps to fine-tune these models by assessing the predictive results through proper feedback mechanisms, re-training on experimental results, and consistently improving the predictive nature of the models.

In addition, the development of localized models based on project-specific datasets will be prioritized to tailor AI models to specific chemical spaces, ensuring that predictions are highly relevant and accurate for specific drug discovery projects. This approach will help in refining the models further, making them more adaptable and precise in predicting ADMET properties across diverse chemical landscapes.

Redefining drug discovery with precision

Aurigene.AI’s ADMET predictive models represent a transformative step forward in drug discovery. As the pharmaceutical industry increasingly adopts AI-driven tools, platforms like Aurigene.AI pave the way for faster, cost-effective, and more successful drug development, bringing life-changing therapies closer to those in need.

Table 1. Illustrative ADMET models present in Aurigene.AI

EndpointScoreMetricADMET classSize of datasetAccuracy logdiff(in %)Confidence Metric at 95%
Lipophilicity0.449 ± 0.009MAEA4,20090.112.251, 2.400
Caco20.285 ± 0.005MAEA906100-5.424, -5.26
Aqueous Solubility0.753 ± 0.004MAEA9,98274.96-3.451, -3.275
Bioavailability0.745 ± 0.005AUROCA6401000.796, 0.887
PGP0.929 ± 0.001AUROCA1,2121000.414, 0.516
HIA0.984 ± 0.004AUROCA5781000.729, 0.864
BBB0.919 ± 0.005AUROCD1,9751000.738, 0.787
VDss0.585 ± 0.0SpearmanD1,13091.590.473, 0.677
CYP2C9 Inhibition0.788 ± 0.002AUPRCM12,0921000.297, 0.321
CYP2D6 Inhibition0.725 ± 0.002AUPRCM13,1301000.154, 0.171
CYP3A4 Inhibition0.882 ± 0.002AUPRCM12,3281000.441, 0.466
CYP2C9 Substrate0.405 ± 0.008AUPRCM6661000.144, 0.181
CYP3A4 Substrate0.657 ± 0.012AUPRCM6671000.538, 0.602
CYP2D6 Substrate0.718 ± 0.002AUROCM6641000.270, 0.346
LD50 0.613 ± 0.025MAET7,38582.402.618, 2.695
hERG0.871 ± 0.003AUROCT6481000.740, 0.824
AMES0.867 ± 0.002AUROCT7,2551000.587, 0.619
DILI0.927± 0.0AUROCT4751000.529, 0.690

About author:

Dr. Lijo has played a pivotal role in developing the AI ( Artificial intelligence) Assisted Drug Discovery Platform | CRO Company platform by closely collaborating with the CADD group and technical team. His efforts in testing and integrating various AI and data-driven models have been crucial in tailoring the platform to meet end-user needs and aligning scientific and technical objectives. He was key in preparing the platform for the BioAsia 2024 event and continues to contribute to its ongoing enhancement. He earned his PhD in 2022 from the CSIR-Indian Institute of Chemical Technology, Hyderabad. He has extensive experience in applying data-driven AI methods for chemical space (structural classification), development of AI/ML methods in small molecule screening, and developing web-based platforms for drug discovery (Hit identification). He has authored 10 publications in peer-reviewed journals.

TAGS

  • Admet
  • Drug Discovery

Social Share

Latest Posts

ADMET Predictive Models in Aurigene

ADMET Predictive Models in Aurigene.AI

January 30, 2025
Arrow sign
Toxicology assessment to accelerate IND and NDA Submissions

Good practices in non-clinical toxicology assessment to accelerate IND and NDA Submissions

October 9, 2024
Arrow sign
PROTACs: Research for a life without cancer

PROTACs: Research for a life without cancer

October 9, 2024
Arrow sign
Continuous Flow Chemistry

Continuous Flow Chemistry: A Game-Changer for Pharmaceutical Production

September 17, 2024
Arrow sign
Why meet us at CPHI Milan?

Why meet us at CPHI Milan?

August 29, 2024
Arrow sign
mRNA Technology and the Brief History of New Vaccine

mRNA Technology and the Brief History of New Vaccine

July 25, 2024
Arrow sign
CDMO white paper

Advancement in personalized medicine and how the CRDMO industry is part of the solution

December 13, 2023
Immuno oncology cdmo

Immuno Oncology - Therapies and Challenges

July 7, 2023
long term partnership with cdmo

Building successful long-term partnerships with CDMOs from early drug discovery through commercialization

December 12, 2022
View All
×

You are about to leave Aurigene Pharmaceutical Services and affiliates website. Aurigene Pharmaceutical Services assumes no responsibility for the information presented on the external website or any further links from such sites. These links are presented to you only as a convenience, and the inclusion of any link does not imply endorsement by Aurigene Pharmaceutical Services.

If you wish to continue to this external website, click Proceed.

ProceedBack