Zeptoward : in silico ADMET property prediction
Discover our next generation AI solution for discovery of absorption, distribution, metabolism, excretion, and toxicity properties
What is prediction of ADMET properties
ADMET properties prediction is the action of predicting how a certain compound will behave in terms of :
Absorption: measures the capacity of a compound to penetrate the bloodstream - sometimes via mucous surfaces like the digestive tract.
Distribution: measures the capacity of a compound to be carried to its effector site.
Metabolism: measures how a drug will perform biotransformation through the action of given organs or tissues so it can be properly excreted.
Excretion: measures how well a metabolized drug compound will be eliminated from the body.
Toxicity: measures if a drug will have a toxic effect on body mechanisms or organs.
Why ADMET property prediction is hard
Several limitations exist in the scoring mechanisms for ADMET:
In vitro methods are often expensive and can only test a single sub-property (e.g. such as testing whether compounds cause a particular type of liver toxicity).
In vivo methods, using animals, bring a host of ethical challenges, can cause unneeded suffering, generate significant costs, and are not always predictive of how drugs will behave in humans.
Most in silico methods have only limited endpoints that they predict, and often have limited predictive performance to new and unseen compounds.
What does zepto.ward do?
Zepto.ward is a machine learning solution (AI) which identifies the ADMET properties of compounds.
It can accurately predict over 80 properties related to absorption, distribution, metabolism, excretion, and toxicity properties, how a specific compound or combination of compounds will perform.
It can also identify off-target effects of compounds.
Because it runs in silico, it can feasibly be run on a full library, rather than only on lead compounds, making it a unique solution to optimize leads and decreasing the risks of drug discovery efforts.
Why is it unique?
Zepto.ward is novel and unique in the following ways :
It can predict ADMET properties across 80 endpoints and enable drug researchers to proactively mitigate unknown effects during the lead optimization phase;
As it is pre-trained on large datasets, it does not need anterior data to generate results.
For the above reason, Zepto.ward can generate results in a matter of minutes;
In the specific case of Targeted Protein Degraders (TPD), Zeptoward can tackle the specific challenge of TPD that do no pass the Lipinski rule.
It can help drug discovery teams optimize the ADMET properties of their compounds.
Last, but not least, it has a very high performance and is therefore a reliable solution to accelerate the work of drug researchers.
What are the benefits
The benefits of using Zepto.ward are the following:
Decrease risk of drug discovery and development by identifying unwanted properties early on;
Increase chances of success during clinical validation;
Decrease costs of lead optimization through a powerful, fast ADMET property analysis;
Speed up time to market by optimizing the ADMET properties early on.
For proper lead optimization, Zepto.ward can be complemented by its sister solution Zepto.hit, which identifies ADMET properties of lead compounds.
Kantify collaborated with I-Stem, the largest French laboratory for research and development dedicated to human pluripotent stem cells. In this project, Kantify and I-Stem used Zepto.ward to perform an artificial intelligence-based predictive ADMET characterization of promising hits, which, combined to a screening performed by I-Stem, led to identification of the HDAC inhibitor givinostat as potential therapeutical candidate.
The publication Disrupting proteasomal and autophagic degradation systems of misfolded alpha-sarcoglycan protein by bortezomib and givinostat combination is available on Authorea.
Interested to learn more?
Read more below about Zeptomics, our AI powered drug discovery pipeline and discover the other possibilities offered by this powerful end-to-end solution.