News

Nature
nature.com > articles > s41375-026-02893-7

Artificial intelligence differentiates prefibrotic primary myelofibrosis with thrombocytosis from essential thrombocythemia using digitized bone marrow biopsy images - Leukemia

3+ hour, 46+ min ago  (1657+ words) Leukemia (2026)Cite this article In this study, we developed an AI model capable of distinguishing prePMF with thrombocytosis from ET on H&E-stained bone marrow biopsy images, with demonstrated potential utility across patient cohorts and institutions. With only knowledge of…...

Nature
nature.com > articles > s42256-026-01182-3

Preconditioned inexact stochastic ADMM for deep models - Nature Machine Intelligence

22+ hour, 25+ min ago  (1670+ words) A preprint version of the article is available at arXiv. In this work, we propose a data-driven preconditioned inexact SADMM algorithm, termed PISA. It distinguishes itself from previous approaches and aims to reduce computational costs, relax convergence assumptions and enhance…...

@nature
nature.com > articles > d41586-026-00557-1

Daily briefing: Automated robot ‘scientists’ spark debate over the future of lab work

1+ day, 14+ hour ago  (273+ words) Chatbot-driven lab robots are automating methods such as protein synthesis.Credit: Du Yu/Xinhua via Alamy Reference: bioRxiv preprint (not peer reviewed) Today, as it rains in the UK for the umpteenth day in a row, I'm once again reaching…...

@nature
nature.com > articles > d41586-026-00365-7

‘An AlphaFold 4’ – scientists marvel at DeepMind drug spin-off’s exclusive new AI

1+ day, 23+ hour ago  (610+ words) The AI tool includes predictions of how proteins interact with potential therapeutic molecules.Credit: Isomorphic Labs Nearly two years after Google DeepMind released an updated AlphaFold3 geared at drug discovery, its biopharmaceuticals spin-off, Isomorphic Labs, announced an even more powerful artificial-intelligence…...

Nature
nature.com > articles > s41586-026-10141-2

Accurate predictions of disordered protein ensembles with STARLING

2+ day, 9+ hour ago  (1670+ words) Model accuracy depends on the balance between ensemble size and denoising steps. By comparing STARLING results to long-timescale Mpipi-GG simulations, we found that error in the Rg stabilizes beyond 30 denoizing steps and 400 conformations (Extended Data Fig. 4c'i). We adopted these as…...

Nature
nature.com > articles > s41388-026-03697-w

Modelling adipose tissue-cancer crosstalk: a three-dimensional perspective - Oncogene

2+ day, 13+ hour ago  (1789+ words) Oncogene (2026)Cite this article This review aims to describe the current knowledge of the interaction between adipose tissue and cancer cells, with a focus on the 3D human models employed to investigate their crosstalk. This work was supported by Action Against…...

Nature
nature.com > articles > s41467-026-68800-x

Chemistry-driven autonomous nanopore membranes - Nature Communications

2+ day, 17+ hour ago  (1644+ words) The high efficiency of Iion spiking allowed the observation of Cl" transport signature through the nanopores. Whereas the log10Ip distribution is featureless for 13,204 Iion spikes detected with a single 100-nm-sized nanopore, a pronounced peak appears ~1.8nA, well described by the Gaussian function…...

@nature
nature.com > articles > d41586-026-00288-3

AI offers way to image and assess clinical cell samples

2+ day, 22+ hour ago  (260+ words) Yinuo Xu is in the Department of Computer and Information Science, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA. Zhi Huang is in the Department of Pathology and Laboratory Medicine, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA. Turner, S. A., Abou Shaar, R. & Yang, Z. Diagn. Cytopathol....

Nature
nature.com > articles > s43246-026-01105-0

AI-powered open-source infrastructure for accelerating materials discovery and advanced manufacturing - Communications Materials

3+ day, 9+ hour ago  (1673+ words) Schematic timeline illustrating the evolution of materials discovery, from empirical and theoretical approaches in the early twentieth century, through computational methods and first-principles modeling in the 1960s1970s, the emergence of data-driven materials science in the 2000s, and the rise of generative AIenabled…...

@nature
nature.com > articles > d41586-026-00495-y

Statistical approximation is not general intelligence

3+ day, 11+ hour ago  (189+ words) Sapienza University of Rome, Rome, Italy. New York University, New York City, New York, USA. In a Comment, Chen et al. argue that success in behavioural tests " including variants of the Turing test " is evidence of artificial general intelligence (AGI;…...