ShakeNBakeGibson

Love that we have one of our first questions even before the official start. Honestly, the millions of simulations that fail enable the one that solves the problem. Both matter!

…and yes, Viagra was a drug originally developed for hypertension and angina pectoris, and as the story goes, when the drug didn’t work that well for those indications and they stopped the trials, none of the participants wanted to give back their clinical trial drugs…. because, well, you know…

But counting on serendipity to give us outcomes like that, in diseases of higher unmet need of course, is not a recipe for success. So we’ve created Recursion to systemize serendipity. But we aren’t stopping at known drugs… we’ve built a dataset spanning over a million molecules that could help us find totally new drugs for many diseases. So its alternative uses, new uses, unexpected uses, and more.

My super fun lawyer would want me to also say: this discussion may contain forward looking statements that are based on current day estimates and operations and importantly are subject to a number of risks. For more details please see the "Risk Factors" in our 10-Q and 10-K SEC filings.

EDIT: added link to comment

We spend a lot of time with investors and analysts in a wide variety of forums from the JP Morgan Healthcare conference to social media. For example, we recently spent a whole day with our analysts and many key investors digging deep into our strategy, platform, pipeline and partnerships at [Download Day](https://www.recursion.com/download-day). You can watch all four hours of detailed content, including questions from analysts at the link.
We think spending <1% of our time finding creative ways to connect to new audiences is a good use of time. We know there are potential future employees on reddit, potential partners and collaborators and more on here. And if we can inspire a bunch of 14 year olds to use their talents for science, that sounds like a win too.

All reductions of complex biology cut out some of the information and become poorer representations of the patient. Scale and translation are opposing forces in biological experimentation. The most translational model is human - which is hardest to scale. The least translational model is in silico, but is easiest to scale.

What we do at Recursion is work in a human cell, the smallest unit of biology that has all of the instructions. It is not perfectly translational, but there are many examples of where it has worked well. But it does allow us to scale across biology and chemistry (whole genome scale, ~1M compounds, etc).

Using that model, we find the strong correlates of gene function and patient biology from the world’s knowledge of disease, and explore those in our dataset to find ways of modifying those processes. We then do the rigorous work of translating success from our cellular models in much more complex systems. Our clinical programs demonstrate that we are able to confirm these insights from the platform in more complex in vivo models.

I love this question. We’re really lucky to already be working with two dream partners! One with Bayer in fibrosis and one with Roche/Genentech in neuroscience and a single oncology indication.

What we look for in new, transformational partnerships are threefold:

1. Learning for us - can we learn from a partner to make the company better for the future?
2. Impact - can we drive value for patients and our shareholders?
3. Data - can we gain access to, retain access to, subsidize access to, or otherwise build our dataset?
   

[Edited - list formatting]

Thank you for the questions!
AI has made huge inroads into tough problems like protein folding. Huge credit to Deepmind and so many others there!
We’ve gone after a different problem than AlphaFold (and others). Can we understand the function of all the proteins in our body without necessarily needing to know the structure? If one could understand cause and effect of all the proteins (when they are overactive, not present, or broken, etc), we could start to better understand what protein to target… and that is important because 90% of drugs that go into clinical trials fail and most often that is because the wrong target is picked.
In terms of successes predicting the results of experiments — we can test ourselves by looking for “ground truths” about biology and chemistry – relationships and pathways that have been proven out in humans – that show up in our maps of biology and chemistry. When our teams search the map and see landmarks they expect, it gives them (and us) extra confidence to explore new ideas surfaced there.
And to your final question – while I can’t say exactly what we’ll charge for future medicines because we’re still fairly early in the development process, I do believe the best way to bring down drug prices is to industrialize the drug discovery process. If we can find a way to scale our pipeline, bringing better medicines to patients faster, with less failure, we can start to bend the cost curve. That’s our goal in the coming decades.