#ai-designed-proteins

#ai-designed-proteins

Small organic molecules underpin modern life, from medicines and flavours to advanced materials. Much of this functional diversity comes from shape: modest changes in a molecule's 3D structure can completely change its properties.

We asked seven frontier AI models to do a simple task. Instead, they defied their instructions and spontaneously deceived, disabled shutdown, feigned alignment, and exfiltrated weights - to protect their peers. We call this phenomenon 'peer-preservation.'

Using CRISPR-Cas9 and adeno-associated virus (AAV)-mediated homology-directed repair, we targeted CAR integration into the endogenous human TCR alpha locus (TRAC). TRAC-CAR T cells display dynamic CAR expression that delays exhaustion and improves tumour control in xenograft and immunocompetent models. This work has been critical for the development of allogeneic CAR T cell therapy, as it disrupts the TCR after transgene insertion—a necessary step to limit graft-versus-host disease.

Since its release in 2021, this repository has become a bedrock in discovery and a first port of call for research projects that try to understand life at the molecular level. But previous iterations of the database lacked predictions of how proteins form complexes, which can be indispensable for their function.

This is a state where we see that the teams that move fastest will be the ones with clear tests, tight review policies, automated enforcement and reliable merge paths. Those guardrails are what make AI useful. If your systems can automatically catch mistakes, enforce standards, and prove what changed and why, then you can safely let agents do the heavy lifting. If not, you're just accelerating risk,

A dyad has three parts, not two: Partner A, Partner B, and the relationship or agreements between them. A dyad of two experts who cannot communicate clearly will often lose to a dyad of less-skilled individuals who coordinate effectively.

Biology is undergoing a transformation. After centuries of studying life as it evolves naturally, researchers are now using a combination of computation and genome engineering to intervene, generating new proteins and even whole bacteria from scratch. The use of artificial-intelligence tools to design biological components, an approach known as generative biology, is set to turbocharge this area of research. Just last year, scientists used AI-assisted design to produce artificial genes that can be expressed in mammalian cells.

Scientists in the laboratory of Rendong Yang, PhD, associate professor of Urology, have developed a new large language model that can interpret transcriptomic data in cancer cell lines more accurately than conventional approaches, as detailed in a recent study published in Nature Communications. Long-read RNA sequencing technologies have transformed transcriptomics research by detecting complex RNA splicing and gene fusion events that have often been missed by conventional short-read RNA-sequencing methods.

Now, researchers have created an artificial-intelligence system that vastly simplifies and accelerates the process of chemical synthesis. The system, which is called MOSAIC and is described in a study published in Nature on 19 January, recommended conditions that researchers were able to use to generate 35 compounds with the potential to become products like pharmaceuticals, agrochemicals or cosmetics without needing to do any further trawling or tweaking.

Despite successes in replicating the primary-secondary-tertiary structure hierarchy of protein, it remains elusive to synthetically materialize protein functions that are deeply rooted in their chemical, structural and dynamic heterogeneities1,2,3,4,5,6,7,8,9,10,11,12. We propose that for polymers with backbone chemistries different from that of proteins, programming spatial and temporal projections of sidechains at the segmental level can be effective in replicating protein behaviours13,14; and leveraging the rotational freedom of polymer can mitigate deficiencies in monomeric sequence specificity and achieve behaviour uniformity at the ensemble level2,3,15,16,17,18,19,20. Here, guided by the active site analysis of about 1,300 metalloproteins, we design random heteropolymers (RHPs) as enzyme mimics based on one-pot synthesis.

Called AlphaGenome, the model could help scientists discover why subtle differences in our DNA put us at risk of conditions such as high blood pressure, dementia and obesity. It could also dramatically accelerate our understanding of genetic diseases and cancer. The developers of the model acknowledge it's not perfect, but experts have described it as "an incredible feat" and "a major milestone".

Generalist models "fail miserably" at the benchmarks used to measure how AI performs scientific tasks, Alex Zhavoronkov, Insilico's founder and CEO, told Fortune. " You test it five times at the same task, and you can see that it's so far from state of the art...It's basically worse than random. It's complete garbage." Far better are specialist AI models that are trained directly on chemistry or biology data.

Structural biology is essential for understanding diseases and for developing drugs and vaccines. Africa has few specialists in this field, owing to limited infrastructure, training and mentorship opportunities - despite the efforts of non-profit organizations such as BioStruct-Africa, which I co-founded. doi: https://doi.org/10.1038/d41586-026-00072-3Competing Interests E.N. received a 2024 Google Award for a socially impactful project enabled by AlphaFold and Google DeepMind sponsorship in 2025 to support the BioStruct-Africa structural-biology training event (series 6). E.N. is also supported by a Wellcome Trust award (grant number 222999/Z/21/Z).