Neural networks’ anticancer remedies

Andrei Voronkov, co-founder of SONM (Supercomputer Organized by Network Mining) project, launched a medical startup of processing medical bioinformatics and chemical informatics data. At the moment, his Drug Discovery@home strives to create a remedy to cure rectal cancer which will have no side effects. Andrei Voronkov explained the details in an interview to Invest Foresight.

Leaving SONM

SONM is a fog computing project. It is about amassing private computers into a computation system which is capable of resolving any computation problems. SONM was founded by computer expert Sergey Ponomarev and chemist researcher Andrei Voronkov. The project was started to research a molecule for a medicine, but ended up as an idea of a multifeature startup, a distributed general purpose supercomputer. In 2017 SONM raised $42 mio at an ICO, and the sale until now remains among top five Russian crowdfunding ventures.

Voronkov abandoned the project in May 2017 to focus on scientific research. Besides, SONM technology has not yet resolved the safety problem for the data submitted to other computers. Yet without such a solution confidential data can not be processed. Unclassified data processing is also in demand, but the prices here are lower. Andrei Voronkov’s new team is ready to offer a solution to the data safety problem and, depending on the SONM’s advances in this area, to consider a possibility of launching a new startup.

The entrepreneur was paid about one percent of the funds raised at SONM’s ICO and founded a company of his own, Drug Discovery@home. This is a distributed platform for computer modeling of medicines intended for specific biotargets.

New cancer cure

Drug Discovery@home mainly designs new methods of developing medications and tests them for treating cancer. They get orders from various research laboratories. At the same time, the company conducts some research of its own, financed from the funds generated by SONM’s ICO (about $420K). Voronkov hopes to create an anticancer medicine which is more efficient than the present ones but has no serious side effects. Chemical therapy can often save human lives but brings about severe after-effects such as heart troubles, infertility, hair loss, etc. Drug Discovery@home’s method allows developing pharmaceuticals for curing currently incurable health problems, which at the same time cause minor adverse reactions.

At the moment, the startup tests a medicine to treat rectal cancer which is the most common type of cancer nowadays. The tests are performed on a cell level in private clinics. There is still a long way to go to clinic tests therefore the medicine has not got a name yet, just a code. A proper name will be given prior to the product release. Some pharmaceutical companies, including from among the Big Pharma, are willing to buy the patent, but Voronkov is interested in selling the product at its final readiness phase. To continue the medicine development, he considers the options of crowdfunding and venture investments.

The new remedy for rectal cancer is yet at an early stage of its development, but Andrei Voronkov is certain about its success, noting that otherwise he would not have invested his personal money into the medicine development.

Drugs computer modeling

The company aims to develop not a single medicine, but a variety. Potentially, the method which Drug Discovery@home has been developing for the past two years, makes that possible. The method is based on machine learning, computer modeling and new neural networks technologies. First, data is collected from scientific literature on what remedies are required for treatment of specific cancer proteins (biotargets). Then, it is to be researched if 3D structures of such biotargets are available. The next stage is verifying if there exist any remedies for such 3D structures. Further on, two or three biotargets are selected for teaching a neural network find proper new active ingredients among non-medicines or biologically inactive substances. Then, medicines and non-medicines are interfused (making tens of thousands of combinations) and a neural network must choose the right combinations which will ultimately be used to produce a required remedy. The combinations obtained are then verified at a test selection. The program thus learns to model interactions on nuclear and molecular levels to see what chemical entities may be efficient. The result is then checked by a properly taught neural network. The main innovation of the company is the use of neural networks for developing molecular dynamics models, intended for each particular biotarget.

Computer modeling is the initial stage of medicines development. It bears essential risks since at that stage is not clear yet, what kind of product will be ultimately developed. Distributed computing breaks down problems into subproblems which are sent out to many computers to be resolved. If the problem requires not distributed but parallel computing, the very same network can be expanded by a large cluster of computing devices where a problem may be referred.

Andrei Voronkov plans to register a patent for his medicines modeling method and get the market share relating to the preclinical studies of medications.

Big Pharma companies, even though they have large R&D divisions and employ computer modeling methods for medicines development, do not have sufficient resources to cover all niches and therefore small innovative companies can create highly competitive innovative products.

By Natalia Kuznetsova

Previous ArticleNext Article