By massively optimizing drug research, this AI-based platform could lead to the discovery of drugs that save millions of lives
AI has revolutionized the field of drug discovery. By processing huge amounts of data in drastically shorter timeframes than would ever be possible by human researchers alone, discoveries are able to happen much faster.
A computer-aided drug design software, the goal of the DeepDrug pipeline is to rapidly develop new antibiotics, antifungals, and antivirals. Not only can the pipeline develop new drug compounds, but it should be able to repurpose existing FDA-approved drugs and chemicals that are either not in use today or are used for completely different purposes. As we continue to live through a global pandemic and grow increasingly resistant to antibiotics, the DeepDrug team is using their technology to work on breakthrough solutions.
Based in the US, the team is led by Dr. Supratik Mukhopadhyay, a Computer Scientist, and Dr. Michal Brylinski, a Computational Biologist. Below, they tell us more about DeepDrug’s potentially life saving work.
Tell us about your team and your AI for good…
Our team is based in Baton Rouge, Louisiana. But there are members scattered around the country from Los Angeles to South Carolina. We’ve created a time-efficient AI-based platform to discover new drugs (or repurpose existing drugs) that can:
- Combat illnesses with no known cure
- Replace drugs for pathogens that have become drug-resistant
- Respond rapidly to outbreaks of previously unknown diseases
- Improve the health of the world's population.
DeepDrug pipeline uses AI-based techniques to process very large datasets, thereby creating an improved method for identifying new drug compounds rapidly, dramatically shortening the early-stage discovery of new drug compounds from years to months, perhaps even weeks.
What does this mean for COVID-19?
As of January 2021, the United States (U.S.) has more confirmed cases of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) than any other country. At present, a few drugs, such as RemDesivir, have been approved by FDA for off-label use in treating SARS-CoV-2 infections. Most of these proposed treatments have been discovered based on trial and error experiences with the virus from around the world. What is needed is a principled approach to drug discovery and repurposing that can rapidly address large datasets, thereby creating an improved method for identifying drugs and/or drug combinations that are very likely to succeed.
We are using AI techniques to rapidly repurpose existing FDA-approved drugs in a principled way in-silico to identify drugs or drug combination therapies that are very likely to inhibit the viral mechanisms of SARS -CoV-2, easing symptoms, lowering the morbidity and/or mortality rate; thereby providing an effective means for dealing with the current pandemic.
Currently we have discovered several FDA-approved drugs that have shown more efficacy than Remdesivir in in-vitro tests (conducted at Iowa State University and IIT Research Institute (IITRI), Chicago) and have discovered drug combinations that can not only reduce the viral load by preventing entry, fusion, and replication, but can also prevent cytokine storm and other effects of Covid. These drugs and drug combinations are now in the process of undergoing human and animal studies. (For more information about these drugs and drug combinations, please feel free to contact me at [email protected]).
Why are your team’s efforts important now, and how do you see them scaling up in the future?
While there have been several vaccines that are 90-95% efficacious, it will take a long time before the entire world’s population can be vaccinated. So, therapeutics that can save lives after infection are very important. A trial and error approach for repurposing the drugs along with their possible combinations with other drugs would be slow and would not be appropriate in a scenario where accelerating the process by a single day can save 1,000 lives. There is a dire need to develop new pipelines and approaches to discover antiviral agents against novel molecular targets such as SARS-CoV-2.
Based on our experience with the current pandemic, we do not think that this one is the last. The in-house preclinical discovery cost and time for a new drug compound by a pharmaceutical company that will lead to a new drug therapy is $209,522,157.00 (adjusted for inflation) and takes at least 3 years (only about 12% of all drugs developed eventually get approved by FDA; the failed attempts significantly increase the cost and time requirement of preclinical drug discovery).
We expect that adoption of the DeepDrug pipeline by the pharmaceutical industry will reduce the time for pre-clinical drug compound discovery and testing from 3 years to a matter of a few weeks and will thereby reduce the cost for this stage of discovery. This is accomplished by removing the inherent inefficacies of high throughput screening that is now used in the search for new drug compounds. The expected impact that we are trying to achieve is the reduction of disease, the alleviation of debilitating health issues, and the improvement of quality of life through a more efficient and cost-effective method for the development of new drug therapies. We expect that through drug repurposing we will be able to save lives within months.
The DeepDrug pipeline can now not only deliver to pharmaceutical researchers antiviral molecules for Covid-19 in the same way that Tamiflu is used for seasonal flu today), but can also rapidly respond to future outbreaks of either SARS-CoV-2 or emerging viruses/pathogens. Factor in the broader application of pipeline to the full spectrum of uses that it can be put to by industry, academic, and governmental researchers, and the number of lives affected expands exponentially.
How has the AI XPRIZE competition furthered your success? How has it changed you?
The competition has brought serious attention to our technology with accompanying funding and other resources.
Lastly, outside of your work, what's an area of AI that's exciting you right now?
Confidence Aware AI: AI that predicts how trustworthy its output is. AI for future data: Building AI where much of the data will be available in the future. For example, designing artifacts that interact with humans where the human interaction data is available much later after design. AI for the design of lipid nanomaterials for drug transport.