Harnessing computation to radically improve lives.

Our Story

We – Prudencio Tossou, Sebastien Giguere, Therence Bois, and Daniel Cohen – have spent the past decade building at the nexus of chemistry, biology, and computer science, first as graduate students, then as co-founders of Valence Discovery. 

From our roots at Mila, and under the influence and mentorship of Yoshua Bengio, we sought to build a new type of company, one with equal parts technology and biotechnology, seamlessly blending computation with the life sciences to unlock drug discovery’s grandest challenges.

Profoundly inspired by first-movers like Recursion — who were demonstrating the scale of the opportunity at the intersection of biomedicine and machine learning — we launched Valence Discovery, in 2018, with the goal of harnessing computation to improve human health, hoping to one day create tangible patient impact. 

And so, we set to work

Unsatisfied with the existing computational toolkit, we set to work, publishing slivers of our progress in top journals and conferences, giving the world a view into the unrelenting and compounding pace of innovation within Valence Discovery.

Building a reputation for living at the frontier — along the way besting giants like Microsoft and NVIDIA in high-profile machine learning competitions — we deployed our technologies internally against challenging molecular targets, and externally alongside forward-thinking biotechnology and pharmaceutical firms. This provided our team with extensive exposure to different drug discovery contexts, resulting in a much deeper understanding of the intricacies of drug discovery than otherwise possible.

Wanting to catalyze progress within our field, we endeavored to share these learnings with the machine learning community, dedicating our time—in the form of reading groups, blogs, newsletters, and communities—and our technology—in the form of a growing collection of open-source tools and open-science collaborations—all in service of the democratization of machine learning in drug discovery. 

Now, just over 5 years in, we feel tremendously privileged to have had a front row seat to—and helped lead—the dizzying rate of progress in our field. From breakthroughs in molecular modeling, to the creation of massive, interrelatable, fit-for-purpose datasets, to a growing number of potential first-in-class and best-in-class medicines progressing in the clinic, the convergence of technology and the life sciences has not only arrived, it has firmly cemented itself as the default modus operandi of modern biotech.

But it's only just the beginning...

Amidst the most dynamic backdrop of progress machine learning has ever seen, turning even hardened skeptics into fervent believers of AI’s potential as a force for good, Valence Discovery has joined forces with Recursion, together creating Valence Labs: a semi-autonomous research and productization engine within Recursion dedicated to advancing the frontier of deep learning in drug discovery. 

We believe this presents a generational opportunity—enabled by the unprecedented combination of Recursion’s scale, computing power, and industrialized data generation capabilities unconstrained by one-off, rate limiting, manual experimentation—to rewrite the rules of therapeutic discovery. 

Combining the intellectual freedom of academia with the resources and stability of industry, Valence Labs will take a long term view on technology development: acting boldly, leaning into risks, and embracing failure, ultimately trading incremental improvements for the breakthrough advances we hope will redefine our field.

Working at the frontiers of machine learning and therapeutic discovery, we commit to deliver both substance—the fundamental research underlying high-impact publications—and form—the product-oriented libraries, models, and content through which the external community can leverage and engage with our work. To redefine our field, we must stop not at GPT-3, but also deliver drug discovery’s equivalent of ChatGPT. 

As we build, we will remain steadfast in our commitment to open-source and open-science whenever possible. We are deeply committed to the machine learning community, aiming to give back at significant levels of investment for rising talent, academic collaborations, and open-science initiatives, both locally within Mila and beyond. 

There is challenging and exciting work to be done, and we hope that you feel inspired to join us in writing this next chapter, embracing the increasingly-connected and ever-evolving worlds of atoms and bits to redefine therapeutic discovery. 

As for us, we recognize the significance of the opportunity ahead: to join forces with a company and team we have so long admired, to build an organization analogous to DeepMind before AlphaGo, OpenAI before GPT, and to bring unrelenting optimism and purpose to our collective mission of harnessing computation to radically improve lives. 


- Dan, Therence, Prudencio, Seb