Thank you so much for the invitation!! It was such a great pleasure visiting NYU Pathology. I truly enjoyed all the meetings and stimulating discussions, always accompanied by that amazing panaromic East River view! 🤩

In 2020 Jennifer Doudna woke up to the news that she had been awarded the Nobel Prize in Chemistry for her work on CRISPR. Who will be woken up with a surprise call this year? Stay tuned. We will find out in October. Learn more: bit.ly/2QRp4Jp

AI turns routine protein measurements into maps of stability A standard experiment can contain far more biology than its standard analysis reveals. Hydrogen-exchange mass spectrometry is widely used to see which regions of a protein become more or less exposed. But those measurements contain something much richer: the local energetic stability of the protein, amino acid by amino acid. PFNet learns to recover that hidden information. The clever part is that it does not need a huge experimental training set. The authors generate synthetic experiments from the underlying physics, add realistic experimental imperfections, and train the model to work backwards from the measurements to the hidden protein energetics. What previously took tens of hours can now be done in seconds. Applied to the SARS-CoV-2 spike protein, that extra resolution reveals not just where ACE2 binds, but how binding redistributes stability across the protein, stabilizing some regions while destabilizing others. AI can increase the resolution of an experiment without changing the experiment itself. Lu et al., Nature Communications (2026), CC BY 4.0. doi.org/10.1038/s41467-026-7…

The first personalized cancer vaccine to succeed in a late-stage trial was tested at UCSF. Earlier results showed it cut the risk of melanoma returning by nearly half. UCSF’s Adil Daud, MD, says it could work for other cancers too. ucsf.edu/news/2026/08/432491…

At OpenAI, it's felt like our researchers & engineers have been living in the future with Codex. Today, we're bringing that same experience to life science researchers with the Rosalind Workbench.

A 2025 #ScienceReview looked at a different approach to drug discovery that may enable drugging of unconventional targets through stabilization of macromolecular complexes with molecules known as “LOCKTACs.” Learn more: scim.ag/4lAiKmF #ScienceMagArchives

A new pharmacological strategy -- based on engineering protein surfaces -- just achieved major success against pancreatic cancer. It points toward a much broader class of cancer therapies. Yesterday, the FDA approved daraxonrasib (Rasonque), a drug that targets the RAS proteins driving most pancreatic cancers. What makes it especially fascinating is how it works. Rather than simply finding a conventional pocket on RAS, daraxonrasib binds the chaperone protein cyclophilin A and uses it to create a new protein surface. This drug–cyclophilin complex then engages active RAS, forming a three-part complex that blocks RAS interactions with downstream effectors. Remarkably, the complex can also stimulate GTP hydrolysis, further suppressing RAS signaling. This is chemical biology at its most powerful: a small molecule is not merely inhibiting a protein: It is engineering a new protein–protein interaction inside the cell to pharmacologically control a previously “undruggable” protein. The clinical impact is already impressive: in a randomized phase III trial, median survival in metastatic pancreatic cancer increased from 6.7 to 13.2 months compared with standard chemotherapy. 🔷 Perhaps the most exciting aspect is what comes next. If we can pharmacologically engineer new protein surfaces and interactions, rather than being restricted to naturally occurring binding pockets, the universe of druggable proteins—and potentially the range of cancers we can treat—could expand dramatically. A beautiful example of how understanding and engineering proteins can open entirely new therapeutic possibilities.