Summary of Science publications, focusing on contents relevant to AI and life sciences
Volume 393 | Issue 6808 | 16 July 2026 — Summary of Science publications, prioritizing content relevant to AI and life sciences. Every item below is individually verified against its DOI and issue tag.
Research Article · pp. 313–318 · DOI: 10.1126/science.aed6123
From Jennifer Doudna's lab at UC Berkeley's Innovative Genomics Institute: enzyme engineering is hard because there are few ways to improve catalytic activity but many ways to break it. This team combined Meta AI's ESM Inverse Folding model (ESM-IF1) — which works in reverse from AlphaFold, generating sequences that fold into a specified backbone — with evolution-informed residue constraints to design new variants of TnpB, a minimal CRISPR-Cas12-like nuclease. The resulting "SynTnpBs" were screened for gene-editing activity in bacterial, plant, and human cells, and cryo-EM structures of the most active variants revealed a previously undescribed TAM-bound conformation stabilizing the RNA-DNA interface. This is a genuinely new-to-nature enzyme design, not just an optimization of an existing one — a meaningful step for generative protein design applied to genome editing.
Editorial · p. 225 · DOI: 10.1126/science.aek5570
Science's Editor-in-Chief takes direct aim at the "faster, better, cheaper" framing popularized in the aerospace industry and now applied to AI by its most enthusiastic advocates, who argue AI will usher in an era of radical abundance freeing humanity from drudgery. Thorp argues history — and his own observations of AI's effect on the scientific publishing pipeline — tells a more complicated story. A useful counterpoint for any AI-optimism narrative Bioinfo.AI might otherwise lean into.
Research Article · DOI: 10.1126/science.aea3075 · see Perspective by FitzGerald, pp. 244–245: DOI 10.1126/science.aei9816
Aging tissue-resident macrophages (TRMs) show increased prostaglandin E2 (PGE2) signaling, which impairs their ability to clear senescent neutrophils — a failure that drives age-related organ decline in mice. Preventing the age-related increase in TRMs restored mitochondrial function and limited several deleterious aging phenotypes across multiple mouse tissues. A clean mechanistic story identifying PGE2 signaling as a druggable node in organ aging, with the accompanying Perspective ("Tidying up aging organs") providing broader context.
Research Article · pp. 306–312 · DOI: 10.1126/science.aeh9302
Using in vivo voltage imaging in the hippocampus, this study shows that dendritic branches of CA3 pyramidal neurons act as independently functioning computational units, capable of coupling or uncoupling from the neuron's main cell body depending on behavioral state. This adds mechanistic detail to how single neurons perform distributed computation — directly relevant to biologically-inspired neural network architectures and any AI/neuroscience crossover content Bioinfo.AI covers.
Research Article · pp. 265–271 · DOI: 10.1126/science.aee3453
The tribosphenic molar — a tooth combining shearing and crushing function — is considered a key adaptive breakthrough in mammalian evolution, but how species navigate the trade-off between the two functions was unclear. Analyzing 250 species (including an extinct saber-toothed cat and living civets balanced between crushing and slicing), the authors quantify a fundamental performance dichotomy shaping mammalian dental evolution. The cover image shows the comparative skull/jaw morphology behind this finding.
Research Article · DOI: 10.1126/science.adx8675
α-Ketoglutarate (αKG) is a metabolite required for chromatin demethylation, but how its nuclear abundance is controlled has been poorly understood. The team transplanted a bacterial αKG-sensing transcription factor into human cells to build a real-time biosensor for nuclear αKG pools — a new tool for probing metabolite-chromatin crosstalk, useful for anyone tracking synthetic-biology tool development in epigenetics.
In Other Journals · pp. 256–257 · DOI: 10.1126/science.aek5615
Summarizing Glewwe et al. (Neuropsychopharmacology), this piece covers a touch-screen rule-switching task in mice: female mice showed greater cognitive flexibility than males, performing more rule-shifts with fewer errors. Computational modeling attributed this to faster commitment to new rules — paradoxically, enhanced choice stability promoted flexibility. Relevant background for computational/behavioral modeling work.
News
NSF has decided to ban collaborations between every U.S. scientist it funds and nearly all Chinese research institutions, abandoning its earlier attempt to balance risk against scientific benefit. Relevant context for anyone tracking the international research-collaboration landscape, including in AI and genomics.
Issue theme in brief: This is a genuinely strong week for Bioinfo.AI's beat — the Skopintsev/Doudna AI-CRISPR paper is the clearest "AI directly advancing life sciences tooling" story you've had in several issues, and it's paired nicely with Thorp's skeptical editorial on AI in the scientific process itself, giving you both the technical win and the critical counterpoint in the same issue.
Comments (0)
No comments yet. Log in to leave a comment.
You must be logged in to leave a comment.
Login to Comment