Ian Bolland speaks to Epistemic AI founders Stefano Pacifico and David Heeger about how its software can be used in drug development and the race to find a cure for Coronavirus.
AI
Epistemic AI has developed what it calls a Knowledge Discovery Platform which is designed to help those working within biomedical research and development. The software uses a mix of machine learning and deep learning algorithms to find existing links within biomedical knowledge, using data from multiple public and private sources.
For pharma, this means those within the industry and academia can use the platform to better understand the genetic mechanisms behind diseases and also find specific biomarkers which could help advance drug discovery.
Give us a bit of background surrounding this software?
Biomedical researchers struggle with information overload, while attempting to grapple with the vast and rapidly expanding base of biomedical knowledge, including documents (e.g., papers, patents, clinical trials) and databases (e.g., genes, proteins, pathways, drugs, diseases, medical terms).
This is a major pain point for researchers and, with no appropriate solution available, they are forced to use basic search tools and explore manually-curated databases. These tools are suitable for finding documents matching keywords (e.g., a single gene or a published journal paper), but not for acquiring comprehensive knowledge about a topic area or subdomain or for interpreting the results of high throughput biology experiments, such as gene sequencing, protein expression, or screening chemical compounds. We started Epistemic AI to address this problem with a platform that empowers users iteratively:
- Shorten the time to gather information and build comprehensive knowledge maps.
- Surface cross-disciplinary information that can be otherwise difficult to find (real discoveries often come from looking into the white space between disciplines).
- Identify causal hypotheses by finding paths and missing links in their knowledge map. And repeat.
We recently released an early beta version of the platform to help COVID-19 researchers. Anyone can sign up at https://covid.epistemic.ai
How does the time saved using software vs human expedite drug discovery?
Today you can explore only one hypothesis, one path at a time, because it is expensive to do so: you have to switch through multiple databases, keyword searches are difficult to drive with precision; in the end that path might be a dead end. Now, in the same amount of time you can explore 10 different ideas, increasing the ability of finding the right path. Also, by being able to more quickly access the knowledge that matters, it’s possible to avoid wasting time on experiments that would have been doomed from the start.
Is the market in danger of becoming over-crowded with this type of software?
That’s a great question. There are many companies that are using some form of machine learning or AI to service the biomedical world. Many successful ones in the drug discovery area use machine learning models to predict the structure of useful molecules. Others offer more traditional tools to search papers (think of variations on the theme of Pubmed and Google Scholar). We do not see many companies that put the user in the driver’s seat while helping them as “a very diligent research assistant”, as one of our scientist users put it.
Following that, what makes Epistemic different from other software available on the market?
We have developed an innovative, AI-powered and interactive platform for researchers to discover knowledge more quickly and efficiently, saving time and money, while also providing completeness and accuracy to make informed decisions, especially at critical stage-gates. This platform can change the way biomedical investigators work and think, thereby fuelling the next generation of breakthrough discoveries and innovations to improve human health.
Think of it as the Bloomberg Terminal for the life sciences. A Bloomberg terminal gives traders the ability to have all the information at their fingertips: news, stock quotes, earning forecasts, price of commodities, bonds, etc. You can find all the information that's useful for making an investment decision, and then you can act on it (trade). Here we take it one step further: all the information at your fingertips (papers, clinical trials, databases of drugs, genes, pathways, etc.) but then we can use Natural Language Programming (NLP) and deep learning algorithms, extract the information and knowledge that's relevant to you and your project, in a prioritised way; discover new information you had not considered that can prove or disprove your thesis; discover linkages among categories of information that may not be obvious; build a knowledge map in minutes and hours when it would have taken days and weeks using conventional tools. We let the user connect the dots in a fast and efficient way.
In terms of Covid-19 research - what are the challenges of directing researchers to the right data when scientists are still struggling to understand the virus?
There are very many struggles that researchers are facing, from the inability to work in wet labs due to social distancing, to the lack of reagents. But what we perceive to be the biggest challenge is the avalanche of new data and knowledge being generated by the scientific community. Any one researcher is trying to drink from the firehose, and that’s hard! Also, to combat a pandemic there needs to be a lot of interdisciplinary collaboration and it thus requires researchers, scientists, and policymakers to comfortably handle knowledge coming from fields in which they are not experts.
What can scientists achieve with the time-saved from sifting through data and research?
They can spend more time actually running experiments, validating hypotheses, but most of all really performing the high-level reasoning tasks that machines and algorithms cannot automate at a sufficiently high level of quality. Who would not want to have more time to do science, rather than spend it to make their tools work for them?
Who have you managed to work with so far?
While I’d love to share our current collaborations, we have not agreed with our users to do so. I can say that we had people signing up from very high profile institutions in academia, non-profit, and pharma. Additionally, we intend to keep the platform free for academic/non-profit purposes.
How do you think this will further drug research in both the short and long term?
In the short term our hope is to provide even just interesting ideas to the tireless researchers that work on urgent issues such as Covid-19. In the long term, we believe this platform can change the way biomedical investigators work and think. We see our platform as a way to further democratise drug research, by reducing costly errors and increasing the likelihood of success.