Taking a drug from research to market is estimated to cost on average 2.6 billion dollars, and more than 10 years.
Currently, researchers are faced with the challenge of manually examining and testing thousands of different compounds to arrive at 5 or less that make it through to clinical trials and demonstrate anti-disease and drug-like properties.
Hence, we developed Synbiolic, a platform that leverages machine learning to accelerate drug discovery by generating molecules and creating retro-synthesis pathways (lego instructions on how to build a molecule).
Specifically, we are deploying a variational auto-encoder, a type of machine learning model that generates new data, trained on a dataset of 74,000 molecules to generate variations of molecular compounds that exhibit drug like characteristics. The generated compounds are filtered by computing its quantitative estimation of drug-likeness (QED), and only choosing molecules with a high score.
To create retro-synthesis pathways to help facilitate the researchers to synthesize the generated compounds, Synbiolic employs the Monte Carlo tree search algorithm. This model is trained on over 1.2 M reactions gathered from a dataset of US Chemical Reaction Patents.
Our machine learning models are developed using Tensorflow, Rdkit (machine learning library for drug discovery), and are trained on a GTX 1050 Ti.
Using Synbiolic, researchers can greatly reduce inefficiencies associated with drug discovery, being able to create medicine faster and cheaper. Our goal with Synbiolic is to give everyone in the world access to medicine, reducing poverty, disease, and creating a better future for humanity.
What inspired you (or your team)?
When I was younger, my family used to take yearly trips to India, visiting the tons of family we had back there. I’d look forward to this trip for the whole year and be ecstatic as we got closer and closer to December (we’d go on Christmas break). I got off the plane every year, jumping up and down with joy, even though my legs were cramping from sitting down for 15 hours, but as soon as I’d leave the airport and go out onto the streets, it amazed me at how much poverty existed. Thousands and thousands of people living literally on the streets in cardboard houses. This was all so foreign to me, being raised in Canada, and seeing people living their lives in these conditions disturbed me. I found that a HUGE number of these people don’t have access to basic medicine, over 50% of people in India, and over 2 BILLION around the world. This seemed insane to me. I dug deeper into why medicine was so expensive, and I found out that the process for drug discovery took on average 12-15 years, and had $2.6 BILLION in costs associated with it, and that treatment for 90% of diseases don’t even exist.
Through my experiences, I became really passionate and invested in how emerging technologies, specifically artificial intelligence, are being leveraged to tackle some of the biggest problems in the world. I found out that the process is hugely ineffecient and has a ton of steps that AI can be trained to solve. After joining The Knowledge Society, I found Joey and Elias, two others that were passionate about solving this problem, and we decided to start Synbiolic people, and take our shot at solving this problem.