Tao Tu

Many bacterial species form self-organized macroscale patterns through swarming. Despite its extensive genetic tractability, Escherichia coli remains underexplored for robust, applied control of swarming. Here we develop a set of E. coli strains that generate centimeter-scale swarming patterns to spatially record environmental inputs. Specifically, we modulate the expression of swarming-related genes in response to chemical and optical signals, reshaping baseline swarm patterns in analog or binary-like fashions. To decode bacterial patterns across space and time, we develop scalable computational methods incorporating feature extraction, regression, and deep-learning models. Time-lapse imaging reveals that colonies record inputs dynamically, enabling early-stage classification. This work establishes a strategy for spatial information recording in E. coli and expands the toolkit for programming emergent microbial behaviors at macroscopic scales.

Although large language models have shown promise in diagnostic dialogue ¹ , their capabilities for effective management reasoning, including disease progression, therapeutic response and safe medication prescription, have remained underexplored. We have advanced the previously demonstrated diagnostic capabilities of the Articulate Medical Intelligence Explorer (AMIE) ^1–3 using a new large-language-model-based agentic system optimized for multivisit clinical management and dialogue. To ground the reasoning of AMIE in authoritative clinical knowledge, we leveraged the long-context capabilities of Gemini ⁴ , combining in-context retrieval with structured reasoning to align its output with up-to-date clinical practice guidelines and drug formularies. In a randomized, blinded virtual Objective Structured Clinical Examination study, AMIE was compared to 21 primary care physicians (PCPs) across 100 multivisit case scenarios designed to reflect the guidance of the UK National Institute for Health and Care Excellence and BMJ Best Practice guidelines. AMIE was non-inferior to PCPs in management reasoning, as assessed by specialists, and scored better both with respect to preciseness of treatment and investigation, and in terms of its alignment with and grounding in clinical guidelines. To benchmark medication reasoning, we developed RxQA, a multiple-choice question benchmark that was derived from two national drug formularies (from the USA and UK) and validated by board-certified pharmacists. Although AMIE and PCPs both benefited from the ability to access external drug information, AMIE outperformed PCPs on higher-difficulty questions. Although further research will be needed before real-world translation of AMIE, its strong performance across evaluations marks a significant step towards use of conversational artificial intelligence as a tool in disease management.

Scientific discovery is driven by scientists generating hypotheses for complex problems that undergo rigorous experimental validation. To augment this process, we introduce Co-Scientist, a multi-agent artificial intelligence (AI) system built on Gemini for structured scientific thinking and hypothesis generation. Co-Scientist aims to help scientists discover new original knowledge. Conditioned on their research objectives and previous scientific evidence, it formulates demonstrably novel research hypotheses for experimental verification. The system’s design involves agents continuously generating, critiquing and refining hypotheses accelerated by scaling test-time compute. Key contributions include (1) a multi-agent architecture with an asynchronous task execution framework for flexible compute scaling, and (2) a tournament evolution process for self-improving hypotheses generation. Automated evaluations show continued benefits of test-time compute scaling, improving hypothesis quality over time. Although this is a general-purpose system, we focus the validation in three biomedical applications: drug repurposing; novel-target discovery ¹ ; and explaining mechanisms of antimicrobial resistance ² . Specifically, Co-Scientist helped to identify new drug-repurposing candidates and synergistic combination therapies for acute myeloid leukaemia that were validated through in vitro experiments. These real-world validations demonstrate the potential of Co-Scientist to accelerate scientific discovery and usher in an era of AI-empowered scientists.

Real-world clinical practice is inherently multimodal, relying on the synthesis of patient history with visual information such as medical imagery and clinical documents. Although large language models (LLMs) have shown promise in diagnostic dialogue, their evaluation has been largely restricted to text-only interactions, failing to capture the complexity of modern remote care delivery. Here we introduce a multimodal extension of the Articulate Medical Intelligence Explorer (multimodal AMIE), capable of gathering, interpreting and reasoning about multimodal data within a diagnostic conversation. To achieve this, we developed a state-aware dialogue framework that dynamically guides history-taking based on diagnostic uncertainty and evolving patient states, emulating the structured reasoning of experienced clinicians. We evaluated this updated, state-aware version of multimodal AMIE against primary care physicians (PCPs) in a randomized, blinded exploratory study comprising 105 simulated telehealth consultations, which included dermatology photographs, electrocardiograms and clinical documents. As assessed by 18 specialist physicians, multimodal AMIE outperformed PCPs not only in diagnostic accuracy but also in conversation quality, including history-taking and empathy. Specifically, multimodal AMIE demonstrated superior performance on 29 of 32 evaluation axes, including seven of nine metrics that assess multimodal reasoning. These results validate the efficacy of state-aware reasoning in bridging the gap between text and visual information and demonstrate the potential for artificial intelligence (AI) systems to augment clinicians in complex, multimodal diagnostic settings.

Artificial intelligence (AI) has been used in many areas of medicine, and large language models (LLMs) have shown potential utility for various clinical applications. However, to determine if LLMs can accelerate the pace of genetic diagnosis and discovery, we examined whether recently developed LLMs (Med‐PaLM 2 and Gemini) could assist in solving four types of genetic problems with sequentially increasing complexity. First, in response to free‐text input, Med‐PaLM 2 correctly identified murine genes with experimentally verified causative genetic factors for six previously studied murine models of biomedical traits. Second, Med‐PaLM 2 identified a novel causative murine genetic factor for spontaneous hearing loss that was validated using knock‐in mice. Third, we developed a retrieval and grounding pipeline that enabled Gemini 2.5 Pro to analyze large lists of genes, which contained genetic variants that were identified in the genomic sequences of 20 human subjects with hearing loss, and demonstrated that it can assist in identifying causative genetic factors for hearing loss. Fourth, we modified the genetic analysis pipeline to enable Gemini 2.5 Pro without any task‐specific fine‐tuning to identify causative genetic factors for six subjects with rare genetic diseases, which required 14 to 34 different terms to describe their multi‐faceted symptom complexes. These results demonstrate that an AI pipeline can facilitate genetic diagnosis and discovery in mice and humans.

The scarcity of subspecialist medical expertise poses a considerable challenge for healthcare delivery. This issue is particularly acute in cardiology, where timely, accurate management determines outcomes. We explored the potential of Articulate Medical Intelligence Explorer (AMIE), a large language model-based experimental medical artificial intelligence system, to augment clinical decision-making in this challenging context. We conducted a randomized controlled trial comparing large language model-assisted care with the usual care of complex patients suspected of having a genetic cardiomyopathy, and we curated a real-world dataset of complex cases from a subspecialist cardiology practice. Nine participating general cardiologists were provided with access to both clinical text reports and raw diagnostic data—including electrocardiograms, echocardiograms, cardiac magnetic resonance imaging scans and cardiopulmonary exercise testing—and were randomized to manage these cases, either with or without assistance from AMIE. We developed a ten-domain evaluation rubric used by three blinded subspecialists to evaluate the quality of triage, diagnosis and management. In our randomized controlled trial with retrospective patient data, subspecialists favored large language model-assisted responses overall, and for the management plan and diagnostic testing domains, with the remaining domains considered a tie. Overall, subspecialists preferred AMIE-assisted cardiology assessments 46.7% of the time, compared with 32.7% for cardiologists alone ( P = 0.02), with 20.6% rated as a tie. Subspecialists also quantified errors, extra and missing content, reasoning and potential bias. Cardiologists alone had more clinically significant errors (24.3% versus 13.1%, P = 0.033) and more missing content (37.4% versus 17.8%, P = 0.0021) than cardiologists assisted by AMIE. Lastly, cardiologists who used AMIE reported that AMIE helped their assessment more than half the time (57.0%) and saved time in 50.5% of cases.

Liver fibrosis has few treatment options due to the poor quality of the available animal and in vitro models. To address this, a hypothesis generating multi‐agent AI system (AI co‐scientist) is used to assist in re‐purposing drugs for treatment of liver fibrosis and direct their experimental characterization. The anti‐fibrotic efficacy and toxicity of 25 drugs are serially assessed in multi‐lineage human hepatic organoids grown in microwells (i.e., microHOs). Remarkably, three previously characterized anti‐fibrotic drugs and two AI co‐scientist‐recommended drugs that targeted epigenomic modifiers exhibited significant anti‐fibrotic activity and they promoted liver regeneration. Analysis of these five anti‐fibrotic drugs revealed that they all can reduce the generation of activated myofibroblasts and that each drug have unique effects on mesenchymal cells that generated their anti‐fibrotic effects. Since all five of the anti‐fibrotic drugs reduced TGFβ‐induced chromatin structural changes, epigenomic changes play an important role in the pathogenesis of liver fibrosis. One AI co‐scientist recommended drug is an FDA‐approved anti‐cancer treatment (Vorinostat) that reduced TGFβ‐induced chromatin structural changes by 91% and promoted liver parenchymal cell regeneration in microHOs. Hence, the integrated use of AI co‐scientist and this microHO platform identified a potential new generation of liver fibrosis treatments that also promote liver regeneration.

AI systems based on Large Language Models (LLMs) have demonstrated great potential for conducting diagnostic conversations but evaluation has been largely limited to language-only interactions, deviating from the real-world requirements of remote care delivery. Instant messaging platforms can permit clinicians and patients to upload and discuss multimodal medical artifacts seamlessly in conversation, but the ability of LLMs to reason over such data while preserving other attributes of competent diagnostic conversation remains unknown. Here we advance the conversational diagnosis and management performance of the Articulate Medical Intelligence Explorer (AMIE) through a new capability to gather and interpret multimodal medical data, and reason about this precisely during consultations. Leveraging Gemini 2.0 Flash, our system implements a state-aware dialogue framework, where conversation flow is dynamically controlled by intermediate model outputs reflecting patient states and evolving diagnoses. Follow-up questions are strategically directed by uncertainty in such patient states, leading to a more structured multimodal history-taking process which emulates that of experienced clinicians. We compared AMIE to primary care physicians (PCPs) in a randomized, double-blind study of chat based consultations with 25 patient actors in the style of an Objective Structured Clinical Examination (OSCE). We constructed 105 evaluation scenarios, using common artifacts such as smartphone photos of skin conditions, ECG tracings, and PDFs of clinical documents across a diversity of conditions and demographics. We designed a rubric to assess multiple aspects of the multimodal capability of AMIE and PCPs, in addition to other clinically meaningful axes such as history-taking, diagnostic accuracy, management reasoning, communication skills, and empathy. Evaluation by 18 specialists revealed superior performance of AMIE relative to PCPs in handling and reasoning about multimodal data on 7 out of 9 axes while displaying similar superior performance in the non-multimodal metrics, including diagnostic accuracy, on 29 out of 32 axes. These results demonstrate clear progress in multimodal conversational diagnostic AI, although real-world translation necessitates further research.

Liver fibrosis is a severe disease with few treatment options due to the poor quality of the available animal and in vitro models. To address this, we investigated whether a hypothesis generating multi-agent AI system (AI co-scientist) could assist in re-purposing drugs for treatment of liver fibrosis and direct their experimental characterization. A multi-parameter image analysis workflow, which enabled anti-fibrotic efficacy and drug toxicity to be serially assessed in multi-lineage human hepatic organoids grown in microwells (i.e., microHOs), was used to assess the effects of 14 drugs. Remarkably, two of the three AI co-scientist-recommended drugs that targeted epigenomic modifiers exhibited significant anti-fibrotic activity. Analysis of the anti-fibrotic effects of five drugs indicated that two inhibited TGFβ-induced intracellular signaling and three drugs altered TGFβ-induced mesenchymal cell differentiation. Since all five of the anti-fibrotic drugs reduced TGFβ-induced chromatin structural changes, epigenomic changes play an important role in the pathogenesis of liver fibrosis. One AI co-scientist recommended drug is an FDA-approved anti-cancer treatment (Vorinostat) that reduced TGFβ-induced chromatin structural changes by 91% and promoted liver parenchymal cell regeneration in microHOs. Hence, the use of AI co-scientist and this microHO platform identified a potential new generation of liver fibrosis treatments that also promote liver regeneration.