The Polish computational chemistry organization Molecule.one has raised $ 4.6 million to expand its quest to realize a theoretical drug molecule. Its machine learning system predicts the best way to synthesize potentially valuable molecules. This is essential for the development of new drugs and treatments.
Molecule.1 Appeared in Startup Battlefield at Disrupt SF 2019, They explained the difficulties facing the drug discovery industry. theorical It’s a cure, but you can’t actually make a molecule.
The company’s system works when you have an exotic new compound that you want to create for testing in real life, but you don’t know how to create it. After all, these molecules are entirely new to science — why does anyone need to know it when no one has created them? Molecule.one is a regular off-the-shelf chemistry Create a workflow starting with a drug and provide step-by-step instructions using known methods such as A to B…, C, D (which is rarely easy).
The company leverages a great deal of knowledge about machine learning and chemistry to create these processes, but they are going in the opposite direction, as CSO Stanislaw Yastortschibski explained.
“Synthetic plans can be characterized as games,” he said. “In each move in this game, instead of moving pieces on the board, we break the chemical bonds between the two atoms. The purpose of this game is to break down the target molecule into molecules, buy them on the market and target them. We use an algorithm similar to that used by DeepMind to master Go and chess and find synthetic pathways. “
Co-founders Piotr Byrski and Paweł Włodarczyk-Pruszy Oschski say that predicting organic reactions is not an easy task and has put a lot of resources into making the system efficient and verifiable. I will. The theoretical path they create looks plausible, but it still needs to be tested.
Since its debut at Disrupt, the company has gained a lot of customers on its annual contract and rolled out a lot of features on the platform, Byrski said. Włodarczyk-PruszyObjectsski says the efficiency has also improved.
“Our system is mature and we have extended the platform to support planned synthesis of thousands of molecules per hour,” he said. “This feature is very useful when combined with an AI system for drug discovery that produces a huge number of candidate drug molecules. All these improvements will earn the trust of the industry and start collaborating with stakeholders. I was able to.”
Indeed, the problem is one of scaling. Customers start asking about hundreds of thousands of possible therapeutic molecule pathways, rather than a handful. For them, when it comes to manufacturing costs, it is worth first checking if one of the compounds under consideration can be made much easier than another compound with similar effects. It’s hard to say for sure without simulating the entire process, so you can just send the list to Molecule.one and you’ll receive the report in a few days.
Of course, this is all confidential, so teams can’t share customer success (although there are probably some). But they said they, like many biotechnology companies, are doing what they can to support COVID-related treatments.
This also provided an opportunity to test new scaling methods.
“This area is generally very exciting because it can cross new areas of the chemical space, which is very promising in looking for drugs that have not yet been synthesized,” said Byrski. I am.
The funding round was led by Atmos Ventures, and many of the participating investors were individuals, including AME Cloud Ventures, Cherubic Ventures, Firlej Kastory, Inventures, Luminous Ventures, Sunfish Partners, and Bayer executive Sebastian Guth.
The company will use the funds to expand the team and continue to expand overall. There are also plans to open new offices in the United States and Western Europe (based in Poland).
Molecule.one grows its drug synthesis AI platform with a $4.6M seed round – TechCrunch Source link Molecule.one grows its drug synthesis AI platform with a $4.6M seed round – TechCrunch