A new Review highlights the mechanisms of molecules that disrupt translation in an amino acid or mRNA sequence-dependent manner nature.com/articles/s41589-0…

I just published: How do we recognize world-class science? Not by the journal it appeared in What I told early-career researchers about prediction, rigour, failure, priors — and the tyranny of bar graphs. Science predicts. Pseudoscience doesn’t. Twenty-nine years How to science, and how not to What’s a failed experiment? Death by statistics Ditch the bar graphs Get credit for all of it kamounlab.medium.com/how-do-…

Spatiotemporal assembly of the human ribosome science.org/doi/10.1126/scie… @ScienceMagazine @EMBLHeidelberg

UCSF’s Kevan Shokat, PhD, just won the $1M Stephenson Global Prize for research in pancreatic cancer. He cracked KRAS, a cancer driver long considered “undruggable.” Drugs built on his work now treat lung and pancreatic cancer. ow.ly/xVlw50ZRSXh

Systematic mapping of bacteriophage gene essentiality with HIDEN-SEQ dlvr.it/TVdsTj #MicroSky

Super postdoc @hemantj32221650 presenting his very cool studies on Mtb gidB variants at the biennual tRNA conference in Crete.

What if any dataset, any size, any number of dimensions, opened in a browser tab from a link? 🔬🧪💻 Luxar is out today: write it in Python, share it as a link, explore it in any browser. Open source. 🧵 @biohub

In light of the progress in mathematics, we at Edison Scientific and FutureHouse have assembled a set of Millennium Problems for Biology. They are chosen to be very hard to solve but very easy to validate in a simple laboratory environment. Any of these, if solved, would mark a major advance in biotechnology, and most of them would contribute materially towards curing disease. These are, in some sense, the “last reasonable eval” for AI in biology. This was work primarily by @MichaelaThinks and myself, with contributions from many others. Short descriptions below. The full descriptions of the problems with acceptance criteria are at the Bio Millennium Problems website, linked in the next post. Share more if you have ideas. If they meet our criteria, we’ll add them to our list (with attribution and permission).

Nature may have explored only a tiny fraction of the proteins that could actually work. That sounds obvious when you consider the size of protein sequence space. But experimentally demonstrating it is much harder. A paper from Lawrence Livermore National Laboratory tested thousands of natural and new-to-nature proteins across three protein families, asking a surprisingly fundamental question: How much functional sequence space exists beyond the proteins evolution has already shown us? The answer appears to be: a lot. Across all three families, the researchers found many experimentally functional proteins far from known natural orthologs. In total, the study measured fitness for roughly 3,000 natural proteins and 10,000 new-to-nature designs, creating an unusually large experimental map of distant protein sequence space. But the second result may be even more important for protein AI. The experimental fitness landscapes were rugged. Protein language models and evolutionary sequence models could often capture broad signals of function, but they did not consistently reproduce either the local shape or the global trends of the measured landscapes. In some comparisons, simpler approaches such as Potts models and HMMs performed as well as or better than much larger AI models. That creates an interesting tension. Most protein foundation models learn heavily from the proteins that evolution happened to preserve. But the set of proteins that exist in nature is not necessarily the same as the set of proteins that could function. Evolution had historical constraints, ecological constraints, mutational paths and billions of contingent events. It never performed an exhaustive search of sequence space. So perhaps natural protein databases should not be treated as a map of functional protein space. They are a sparse evolutionary archive of it. And if functional proteins occupy regions that nature never sampled, the next frontier of protein design is not simply learning the distribution of existing proteins more accurately. It is learning where that distribution ends — and where functional biology continues beyond it. The proteins we know may be islands. The functional sequence space around them may be an ocean. biorxiv.org/content/10.64898…

Identification of chemical features for improved outer membrane permeation in mycobacteria using machine learning | Nature Microbiology share.google/rBhVqC3fuxSulTv…

It took me a while but I wanted to share the main work from Ana MG Marín @amgm_in the lab, exploring how complete genomes can change our view of Mtb diversity accross scales, from global to within host diversity @CSIC @GVAfisabio @UV_EG @CIBER_ISCIII nature.com/articles/s41467-0…

Apply to do a PhD at @sangerinstitute with me on global genomics of Mycobacterium abscessus bbsrcdtp.lifesci.cam.ac.uk/p… About the scheme: bbsrcdtp.lifesci.cam.ac.uk/a…

Delighted that this story is finally out in its final form! Mycobacterial ribosomes can discriminate against mischarged tRNAs! In @eLife: Ribosomal RNA methylation by GidB modulates discrimination of mischarged tRNA doi.org/10.7554/eLife.102752…

Everyone check out this great paper I read, published yesterday. The creating of messenger-ribosomal RNA (mrRNA). Creating ribosomes capable of translating their own RNA! nature.com/articles/s41586-0…

Assessment of differentially culturable tubercle bacteria assays for the detection of tuberculosis infection in asymptomatic household contacts and the implications for intra-household transmission: a longitudinal cohort study - The Lancet Microbe thelancet.com/journals/lanmi…

The bacterial pH gradient contributes to persistence in Mycobacterium tuberculosis | mBio journals.asm.org/doi/full/10…