In Silico Hit Prediction - ZeptoHit
Scope of the method
- Human health
- Translational - Applied Research
- In silico
Description
- in silico analysis
- hit discovery
- phenotypic activity
- computational modelling
- artificial intelligence
- machine learning
- bioinformatics
ZeptoHit is a technology based on Artificial Intelligence (AI) which accelerates hit discovery. With only a protein sequence as information, ZeptoHit can quickly and efficiently recommend promising hits, without any limitations. Its search for promising compounds goes much beyond chemical structures similar to known hits. Furthermore, ZeptoHit can accurately make predictions for proteins that have no known modulator or binder. ZeptoHit can be seen as a very smart filter which will predict promising compounds across millions of possibilities. These compounds can later be screened and validated in a wet lab in a low throughput screening campaign.
- Internally validated
- Published in peer reviewed journal
Pros, cons & Future potential
ZeptoHit has been learning from millions of examples of hits. With only a protein sequence as input, it can generate a list of likely hits, with a high accuracy. Thanks to this wide knowledge it is capable of:
- - identifying not only the on-target hits, but also the off-target hits so that adverse effects can better be anticipated;
- - identifying hits even in the case of unknown compounds;
- - drug discovery teams working on novel compounds do not start from a blank page but can already start identifying possible hits;
- - identifying hits even in the case of novel targets. Also here, pharma researchers can pre-screen a novel target even if it is not well known or studied;
- - predicting the type of bioactivity: the solution does not only predict if there will be a hit but also predicts the type of expected bioactivity: activation, inhibition…;
- - prioritizing compounds that are synthesizable and available so they can be rapidly tested.
The benefits of using ZeptoHit are the following:
- - Screen only the compounds with a likely promising bioactivity on humans so to reduce unnecessary costs and animal testing efforts;
- - Considerably decrease time and cost of drug discovery efforts;
- - Increase chances of success through the identification of possible adverse effects due to off target hits;
- - Filter out non promising leads in order to concentrate screening efforts on promising leads;
- - Decrease risk of drug development projects by having the widest possible search of promising leads, instead of missing out on some valuable compounds.
The method is probabilistic, even if the hit rate is considerably higher than the state of the art. This means that we always work with partners who have mechanistic or phenotypic assays to validate the predictions of ZeptoHit.
ZeptoHit is constantly modified so its predictions constantly improve. Also, it is developed in parallel with ZeptoWard, our ADMETox prediction model, so we can select, from the project start, not only promising compounds, but also safe ones, so animal testing is not performed in vain.
References, associated documents and other information
https://www.frontiersin.org/articles/10.3389/fphar.2022.856804/full
Contact person
Segolene MartinOrganisations
KantifyDrug Discovery Department
Belgium