Audio long read: AI has dreamt up a blizzard of new proteins. Do any of them actually work?
Nature Podcast
Deep Dive
Why are AI-designed proteins being created faster than they can be tested?
AI tools like AlphaFold and protein language models have become more powerful and popular, leading to a surge in the generation of designer proteins. However, the pace of creation outstrips the ability to test these proteins in labs, causing confusion about their effectiveness.
What role do protein design competitions play in the field of protein engineering?
Protein design competitions aim to identify functional proteins from the vast number of designs generated by AI. They lower the barrier to entry for newcomers, accelerate validation and standards development, and foster a sense of community among researchers.
What challenges do protein design competitions face?
Key challenges include identifying relevant problems to tackle, selecting winners objectively, and ensuring that competitions do not narrow the field too much or discourage sharing of methods. Proper execution is crucial to avoid damaging the field.
How did the CASP competition influence the development of AI in protein structure prediction?
The CASP competition, launched in 1994, revolutionized protein structure prediction by challenging scientists to predict 3D shapes from amino acid sequences. It directly led to the creation of AlphaFold, which solved the problem of predicting simple protein structures.
What are some examples of recent protein design competitions?
Examples include the Protein Engineering Tournament by Align to Innovate, the Winter Protein Design Games by Liberum Bio and Rosetta Commons, and the Adaptive Bio competition, which focused on designing proteins to target EGFR in cancer.
Who participated in the protein design competitions, and what were the outcomes?
Participants ranged from professional protein engineers to hobbyists with no formal biology experience. Despite high participation, success rates were low, with only a small percentage of designs proving functional in lab tests.
What are the potential benefits of protein design competitions for the scientific community?
Competitions can accelerate the pace of validation, encourage collaboration, and bring together diverse fields like biochemistry and machine learning. They also provide opportunities for learning and community-building within the protein design field.
What was the outcome of the Adaptive Bio competition, and how did it impact participants?
Only 5 out of 147 tested designs successfully bound to the target molecule, reflecting the challenges of protein engineering. Despite the low success rate, participants gained valuable experience and connections, with some continuing to participate in future competitions.
- AI tools like AlphaFold and protein language models are used in protein design competitions.
- Competitions help to identify functional proteins among many.
- The competitions aim to quicken the pace of validation and standards development.
Shownotes Transcript