Thursday, October 17, 2024
Introducing our Small Molecule Steering Committee
Over the past few years, the field of machine learning for drug discovery (MLDD) has experienced an unprecedented surge of innovation. Breakthrough algorithms and increasing computational power are enabling researchers to tackle complex biological challenges like protein design, molecular generation, target identification, and much more.
For the first time, we saw AI get recognized for its impact on science with the Nobel Prize in Chemistry 2024 being awarded to the AlphaFold team and David Baker’s work on protein design. This breakthrough was enabled by initiatives like CASP, highlighting the importance of well-scoped benchmarks and competitions when it comes to driving innovation in AI for science.
But how do we know whether these innovations can truly impact the development of new drugs? It's hard to tell. The challenge with benchmarking in MLDD is that it’s complex and nuanced. The applications of ML in drug discovery are numerous, require familiarity with several scientific disciplines, and inform high-stakes decisions, such as expensive or time-consuming experiments. The absence of standardized, domain-appropriate datasets, guidelines and tools for the evaluation and comparison of methods has led to a growing gap between perceived progress and real-world impact, which is delaying the adoption of ML in drug discovery.
This is the challenge that we set out to solve with Polaris.
Meet the Steering Committee
To address these challenges, we are forming steering committees comprised of industry experts to craft specific guidelines for the community. By drawing on their practical insights, we aim to develop resources that reflect the complexities of the field and can guide the community toward the development of more impactful ML methods.
Over the past year, our focus has been on small molecules and today we’re excited to introduce the first steering committee:
- Pat Walters - Chief Data Officer, Relay Therapeutics
- Alan Cheng - Senior Director, Merck
- Djork-Arne (Okko) Clevert - Vice President, Pfizer
- Cheng Fang - Associate Director, Blueprint Medicines
- Dan Price - Vice President, Nimbus Therapeutics
- Ola Engkvist - Executive Director, AstraZeneca
- Jeremy Ash - Senior Scientist, Johnson & Johnson
- Matteo Aldeghi - Director, Bayer
- Raquel Rodríguez Pérez - Associate Director, Novartis
- Cas Wognum - Senior Scientist, Valence Labs (powered by Recursion Pharma)
On the Roadmap
This collaboration marks the beginning of a series of efforts aimed at improving machine learning benchmarking for drug discovery. The first correspondence letter published in Nature Machine Intelligence served as a call to action for the drug discovery and machine learning communities. We emphasize the importance of cross-industry, interdisciplinary collaboration to address three main challenges:
- Creating a recommended set of benchmark datasets that represent tasks typically performed in drug discovery.
- A set of guidelines for method comparison (such as statistical tests), method evaluation (such as dataset splitting and evaluation metrics), and dataset curation.
- Creating the open source platform and tools to simplify the adoption of best practices, centred around the Polaris Hub.
To date, we’ve been focused on improving the versatility of Polaris as a benchmarking platform and introducing tools like Auroris to help with dataset curation. With the collective efforts of the steering committee, we’re excited to release the first guideline preprint on method comparison in November.
You’ll be able to find all of the guidelines on Polaris. Stay updated with our latest release by signing up for our mailing list.
Beyond Small Molecules
While our initial focus is on small molecules, we plan to establish a steering committee for other drug discovery-relevant modalities in the future. We invite members of the scientific and industrial communities to join us on this journey. If you are interested in contributing your expertise or have suggestions for how we can enhance our efforts, please reach out.