-
Reconfigurable system for automated optimization of diverse chemical reactions - Science, 2018. [All Versions]. [Preprint]. This paper describes a plug-and-play, continuous-flow chemical synthesis system that mitigates this challenge with an integrated combination of hardware, software, and analytics. The system software controls the user-selected reagents and unit operations (reactors and separators), processes reaction analytics (high-performance liquid chromatography, mass spectrometry, vibrational spectroscopy), and conducts automated optimizations.
-
A universal system for digitization and automatic execution of the chemical synthesis literature - Science, 2020. [All Versions]. [Preprint]. [XDL Documentation]. [XDL Schema Database]. This paper reports a software platform that uses natural language processing to translate the organic chemistry literature directly into editable code, which in turn can be compiled to drive automated synthesis of the compound in the laboratory.
-
Digitization and validation of a chemical synthesis literature database in the ChemPU - Science, 2022. [All Versions]. [Preprint]. This paper presents an automatically executable chemical reaction database of 100 molecules representative of the range of reactions found in contemporary organic synthesis. The chemical reaction codes or χDLs for the reactions have been stored in a database for version control, validation, collaboration, and data mining. Of these syntheses, more than 50 entries from the database have been downloaded and robotically run in seven modular chemputers with yields and purities comparable to those achieved by an expert chemist.
-
Chemputation and the Standardization of Chemical Informatics - Journal of the American Chemical Society (Au), 2021. [All Versions]. This paper describes a standard hardware (the chemical processing programming architecture --- the ChemPU) to encompass all chemical synthesis, an approach which unifies all chemistry automation strategies, from solid-phase peptide synthesis, to HTE flow chemistry platforms, while at the same time establishing a publication standard so that researchers can exchange chemical code (χDL) to ensure reproducibility and interoperability.
-
An autonomous portable platform for universal chemical synthesis - Nature Chemistry, 2022. [All Versions]. [Preprint]. This paper presents a portable suitcase-sized chemical synthesis platform containing all the modules required for synthesis and purification. The system uses a chemical programming language coupled to a digital reactor generator to produce reactors and executable protocols based on text-based literature syntheses. Simultaneously, the platform generates a reaction pressure fingerprint, used to monitor processes within the reactors and remotely perform a protocol quality control.
-
A mobile robotic chemist - Nature, 2020. [All Versions]. [Preprint]. This work uses a mobile robot to search for improved photocatalysts for hydrogen production from water. The robot operated autonomously over eight days, performing 688 experiments within a ten-variable experimental space, driven by a batched Bayesian search algorithm. This autonomous search identified photocatalyst mixtures that were six times more active than the initial formulations, selecting beneficial components and deselecting negative ones.
-
An autonomous laboratory for the accelerated synthesis of novel materials - Nature, 2023. [All Versions]. This paper introduces the A-Lab, an autonomous laboratory for the solid-state synthesis of inorganic powders. This platform uses computations, historical data from the literature, machine learning (ML) and active learning to plan and interpret the outcomes of experiments performed using robotics. Over 17 days of continuous operation, the A-Lab realized 41 novel compounds from a set of 58 targets including a variety of oxides and phosphates that were identified using large-scale ab initio phase-stability data from the Materials Project and Google DeepMind.
-
The Internet of Things comes to the lab - Nature, 2017. [All Versions]. The emergence of connected instruments and equipment promises to untether researchers from the laboratory --- letting them fine-tune experiments and analyse data remotely.
-
A dynamic knowledge graph approach to distributed self-driving laboratories - Nature Communications, 2024. [All Versions]. This work employs ontologies to capture data and material flows in design-make-test-analyse cycles, utilising autonomous agents as executable knowledge components to carry out the experimentation workflow. Data provenance is recorded to ensure its findability, accessibility, interoperability, and reusability. The architecture is built upon the World Avatar project, which seeks to create an all-encompassing digital twin based on a dynamic knowledge graph.
-
Automation isn't automatic - Chemical Science, 2021. [All Versions]. This perspective provides an overview of the current state of automation of synthetic chemistry at the benchtop scale with a particular emphasis on core considerations and the ensuing challenges of deploying a system. The authors aim to reframe automation as decidedly not automatic but rather an iterative process that involves a series of careful decisions (both human and computational) and constant adjustment.
-
Balancing act: when to flex and when to stay fixed - Trends in Chemistry, 2023. [All Versions]. This perspective article provides essential insights into the decision-making process for choosing automation platforms, highlighting the suitability of fixed automation for standardized tasks and the strategic use of flexible automation in dynamic research settings.
-
What is a minimal working example for a self-driving laboratory? - Matter, 2022. [All Versions]. This paper proposes SDL-Demo: a low-cost “Hello, World!” for self-driving laboratories that combines “Hello, World!” tasks from electronics, physics-based simulations, and optimization. SDL-Demo is modular and extensible, making it an ideal candidate for low-cost teaching and prototyping of self-driving laboratory concepts.
-
Robotic search for optimal cell culture in regenerative medicine - eLife, 2022. [All Versions]. This paper develops a robotic AI system with a batch Bayesian optimization algorithm that autonomously induces the differentiation of induced pluripotent stem cell-derived retinal pigment epithelial (iPSC-RPE) cells. From 200 million possible parameter combinations, the system performed cell culture in 143 different conditions in 111 days, resulting in 88% better iPSC-RPE production than that obtained by the pre-optimized culture in terms of the pigmentation scores.
-
Balancing autonomy and expertise in autonomous synthesis laboratories - Nature Computational Science, 2025. [All Versions]. Autonomous synthesis laboratories promise to streamline the plan–make–measure–analyze iteration loop. Here, the authors comment on the barriers in the field, the promise of a human on-the-loop approach, and strategies for optimizing accessibility, accuracy, and efficiency of autonomous laboratories.
-
AlphaFlow: autonomous discovery and optimization of multi-step chemistry using a self-driven fluidic lab guided by reinforcement learning - Nature Communications, 2023. [All Versions]. Closed-loop, autonomous experimentation enables accelerated and material-efficient exploration of large reaction spaces without the need for user intervention. However, autonomous exploration of advanced materials with complex, multi-step processes and data sparse environments remains a challenge. In this work, the authors present AlphaFlow, a self-driven fluidic lab capable of autonomous discovery of complex multi-step chemistries. AlphaFlow uses reinforcement learning integrated with a modular microdroplet reactor capable of performing reaction steps with variable sequence, phase separation, washing, and continuous in-situ spectral monitoring. To demonstrate the power of reinforcement learning toward high dimensionality multi-step chemistries, the authors use AlphaFlow to discover and optimize synthetic routes for shell-growth of core-shell semiconductor nanoparticles, inspired by colloidal atomic layer deposition (cALD). Without prior knowledge of conventional cALD parameters, AlphaFlow successfully identified and optimized a novel multi-step reaction route, with up to 40 parameters, that outperformed conventional sequences. Through this work, the authors demonstrate the capabilities of closed-loop, reinforcement learning-guided systems in exploring and solving challenges in multi-step nanoparticle syntheses, while relying solely on in-house generated data from a miniaturized microfluidic platform. Further application of AlphaFlow in multi-step chemistries beyond cALD can lead to accelerated fundamental knowledge generation as well as synthetic route discoveries and optimization.
-
Scientific discovery in the age of artificial intelligence - Nature, 2023. [All Versions]. A review article that examines breakthroughs over the past decade that include self-supervised learning, which allows models to be trained on vast amounts of unlabelled data, and geometric deep learning, which leverages knowledge about the structure of scientific data to enhance model accuracy and efficiency.
-
Artificial Intelligence for Retrosynthetic Planning Needs Both Data and Expert Knowledge - Journal of the American Chemical Society, 2024. [All Versions]. The development of AI synthesis planners trained solely on reaction-example-data has stagnated and is not on par with the performance of “hybrid” algorithms combining AI with expert knowledge. This Perspective examines possible causes of these shortcomings, extending beyond the established reasoning of insufficient quantities of reaction data. Drawing attention to the intricacies and data biases that are specific to the domain of synthetic chemistry, the authors advocate augmenting the unique capabilities of AI with the knowledge base and the reasoning strategies of domain experts. By actively involving synthetic chemists, who are the end users of any synthesis planning software, into the development process, the authors envision to bridge the gap between computer algorithms and the intricate nature of chemical synthesis.
-
The Impact of Large Language Models on Scientific Discovery: a Preliminary Study using GPT-4 - Microsoft Research AI4Science, 2023. [All Versions]. [Project]. A survey on the performance of LLMs within the context of scientific discovery, focusing on GPT-4.
-
An agentic system for rare disease diagnosis with traceable reasoning - Nature, 2026. [All Versions]. Rare diseases affect more than 300 million people worldwide, yet timely and accurate diagnosis remains an urgent challenge. Patients often endure a prolonged ‘diagnostic odyssey’ exceeding 5 years, marked by repeated referrals, misdiagnoses and unnecessary interventions, leading to delayed treatment and substantial emotional and economic burden. This work presents DeepRare---a multi-agent system for rare disease differential diagnosis decision support powered by large language models, integrating more than 40 specialized tools and up-to-date knowledge sources. DeepRare processes heterogeneous clinical inputs, including free-text descriptions, structured human phenotype ontology terms and genetic testing results to generate ranked diagnostic hypotheses with transparent reasoning linked to verifiable medical evidence. Evaluated across nine datasets from literature, case reports and clinical centres across Asia, North America and Europe spanning 14 medical specialties, DeepRare demonstrates exceptional performance on 2,919 diseases. In human-phenotype-ontology-based tasks, it achieves an average Recall@1 of 57.18%, outperforming the next best method by 23.79%; in multi-modal tests, it reaches 69.1% compared with Exomiser’s 55.9% on 168 cases. Expert review achieved 95.4% agreement on its reasoning chains, confirming their validity and traceability. This work not only advances rare disease diagnosis but also demonstrates how the latest powerful large-language-model-driven agentic systems can reshape current clinical workflows.
-
Towards end-to-end automation of AI research - Nature, 2026. [All Versions]. The automation of science is a long-standing ambition in artificial intelligence (AI) research. Although the community has made substantial progress in automating individual components of the scientific process, a system that autonomously navigates the entire research life cycle---from conception to publication---has remained out of reach. This work presents a pipeline for automating the entire scientific process end to end. The authors present The AI Scientist, which creates research ideas, writes code, runs experiments, plots and analyses data, writes the entire scientific manuscript, and performs its own peer review. Its ideas, execution and presentation are of sufficient quality that the manuscript generated by this AI system passed the first round of peer review for a workshop of a top-tier machine learning conference. The workshop had an acceptance rate of 70%. The system leverages modern foundation models within a complex agentic system. The authors evaluate The AI Scientist in two settings: a focused mode using human-provided code templates as an initial scaffold for conducting research on a specific topic and a template-free, open-ended mode that leverages agentic search for wider scientific exploration. Both settings produce diverse ideas and automatically test, report on and evaluate them. This achievement demonstrates the growing capacity of AI for making scientific contributions and signifies a potential paradigm shift in how research is conducted. As with any impactful new technology, there could be important risks, including taxing overwhelmed review systems and adding noise to the scientific literature. However, if developed responsibly, such autonomous systems could greatly accelerate scientific discovery.
-
Machine learning-assisted molecular design and efficiency prediction for high-performance organic photovoltaic materials - Science Advances, 2019. [All Versions]. In the process of finding high-performance materials for organic photovoltaics (OPVs), it is meaningful if one can establish the relationship between chemical structures and photovoltaic properties even before synthesizing them. This work first establishes a database containing over 1700 donor materials reported in the literature. Through supervised learning, our machine learning (ML) models can build up the structure-property relationship and, thus, implement fast screening of OPV materials. The authors explore several expressions for molecule structures, i.e., images, ASCII strings, descriptors, and fingerprints, as inputs for various ML algorithms. It is found that fingerprints with length over 1000 bits can obtain high prediction accuracy. The reliability of the approach is further verified by screening 10 newly designed donor materials. Good consistency between model predictions and experimental outcomes is obtained. The result indicates that ML is a powerful tool to prescreen new OPV materials, thus accelerating the development of the OPV field.
-
Design of metalloproteins and novel protein folds using variational autoencoders - Scientific Reports, 2018. [All Versions]. The design of novel proteins has many applications but remains an attritional process with success in isolated cases. Meanwhile, deep learning technologies have exploded in popularity in recent years and are increasingly applicable to biology due to the rise in available data. This work attempts to link protein design and deep learning by using variational autoencoders to generate protein sequences conditioned on desired properties. Potential copper and calcium binding sites are added to non-metal binding proteins without human intervention and compared to a hidden Markov model. In another use case, a grammar of protein structures is developed and used to produce sequences for a novel protein topology. One candidate structure is found to be stable by molecular dynamics simulation. The ability of the model to confine the vast search space of protein sequences and to scale easily has the potential to assist in a variety of protein design tasks.
-
Highly accurate protein structure prediction with AlphaFold - Nature, 2021. [All Versions]. This paper provides the first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known. This approach is a canonical application of observation- and explanation- based method for protein structure prediction instead of first-principle-based methods.
-
Human–machine collaboration for improving semiconductor process development - Nature, 2023. [All Versions]. [Nature News]. This work studies Bayesian optimization algorithms to investigate how artificial intelligence (AI) might decrease the cost of developing complex semiconductor chip processes. In particular, this work create a controlled virtual process game to systematically benchmark the performance of humans and computers for the design of a semiconductor fabrication process. The authors find that human engineers excel in the early stages of development, whereas the algorithms are far more cost-efficient near the tight tolerances of the target.
-
A foundation model for generalizable disease detection from retinal images - Nature, 2023. [All Versions]. This paper presents RETFound, a foundation model for retinal images that learns generalizable representations from unlabelled retinal images and provides a basis for label-efficient model adaptation in several applications. Specifically, RETFound is trained on 1.6 million unlabelled retinal images by means of self-supervised learning and then adapted to disease detection tasks with explicit labels.
-
Accurate medium-range global weather forecasting with 3D neural networks - Nature, 2023. [All Versions]. This paer introduces an artificial-intelligence-based method for accurate, medium-range global weather forecasting. It shows that three-dimensional deep networks equipped with Earth-specific priors are effective at dealing with complex patterns in weather data, and that a hierarchical temporal aggregation strategy reduces accumulation errors in medium-range forecasting. Trained on 39 years of global data, the program, Pangu-Weather, obtains stronger deterministic forecast results on reanalysis data in all tested variables when compared with the world’s best NWP system, the operational integrated forecasting system of the European Centre for Medium-Range Weather Forecasts.
-
Learning skillful medium-range global weather forecasting - Science, 2023. [All Versions].
-
Skilful nowcasting of extreme precipitation with NowcastNet - Nature, 2023. [All Versions].
-
Autonomous chemical research with large language models - Nature, 2023. [All Versions]. An artificial intelligence system driven by GPT-4 that autonomously designs, plans and performs complex experiments by incorporating large language models empowered by tools such as internet and documentation search, code execution and experimental automation.
-
Augmenting large language models with chemistry tools - Nature Machine Intelligence, 2023. [All Versions]. [Preprint]. This paper introduces ChemCrow, an LLM chemistry agent designed to accomplish tasks across organic synthesis, drug discovery and materials design. By integrating 18 expert-designed tools and using GPT-4 as the LLM, ChemCrow augments the LLM performance in chemistry, and new capabilities emerge. The agent autonomously planned and executed the syntheses of an insect repellent and three organocatalysts and guided the discovery of a novel chromophore.
-
Empowering biomedical discovery with AI agents - Cell, 2024. [All Versions]. The authors envision “AI scientists” as systems capable of skeptical learning and reasoning that empower biomedical research through collaborative agents that integrate AI models and biomedical tools with experimental platforms. Rather than taking humans out of the discovery process, biomedical AI agents combine human creativity and expertise with AI’s ability to analyze large datasets, navigate hypothesis spaces, and execute repetitive tasks. AI agents are poised to be proficient in various tasks, planning discovery workflows and performing self-assessment to identify and mitigate gaps in their knowledge. These agents use large language models and generative models to feature structured memory for continual learning and use machine learning tools to incorporate scientific knowledge, biological principles, and theories. AI agents can impact areas ranging from virtual cell simulation, programmable control of phenotypes, and the design of cellular circuits to developing new therapies.
-
DrBioRight 2.0: an LLM-powered bioinformatics chatbot for large-scale cancer functional proteomics analysis - Nature Communications, 2025. [All Versions]. [Project]. Functional proteomics provides critical insights into cancer mechanisms, facilitating the discovery of novel biomarkers and therapeutic targets. The authors have developed a comprehensive cancer functional proteomics resource using reverse phase protein arrays, incorporating data from nearly 8000 patient samples from The Cancer Genome Atlas and approximately 900 samples from the Cancer Cell Line Encyclopedia. The dataset includes a curated panel of nearly 500 high-quality antibodies, covering all major cancer hallmark pathways. To enhance the accessibility and analytic power of this resource, this work introduces DrBioRight 2.0, an intuitive bioinformatic platform powered by state-of-the-art large language models. DrBioRight enables researchers to explore protein-centric cancer omics data, perform advanced analyses, visualize results, and engage in interactive discussions using natural language. By streamlining complex proteogenomic analyses, this tool accelerates the translation of large-scale functional proteomics data into meaningful biomedical insights.
-
The Virtual Lab of AI agents designs new SARS-CoV-2 nanobodies - Nature, 2025. [All Versions]. Science frequently benefits from teams of interdisciplinary researchers, but many scientists do not have easy access to experts from multiple fields. Although large language models (LLMs) have shown an impressive ability to aid researchers across diverse domains, their uses have been largely limited to answering specific scientific questions rather than performing open-ended research. Here the authors expand the capabilities of LLMs for science by introducing the Virtual Lab, an artificial intelligence (AI)–human research collaboration to perform sophisticated, interdisciplinary science research. The Virtual Lab consists of an LLM Principal Investigator agent guiding a team of LLM scientist agents through a series of research meetings, with a human researcher providing high-level feedback. The authors applied the Virtual Lab to design nanobody binders to recent variants of SARS-CoV-2. The Virtual Lab created a novel computational nanobody design pipeline that incorporates the protein language model ESM, the protein folding model AlphaFold-Multimer and the computational biology software Rosetta and designed 92 new nanobodies. Experimental validation reveals a range of functional nanobodies with promising binding profiles across SARS-CoV-2 variants. In particular, two new nanobodies exhibit improved binding to the recent JN.1 or KP.3 variants while maintaining strong binding to the ancestral viral spike protein, suggesting that these are suitable candidates for further investigation. This work demonstrates how the Virtual Lab can rapidly make an impactful, real-world scientific discovery.
-
BioPlanner: Automatic Evaluation of LLMs on Protocol Planning in Biology - EMNLP'23, 2023. [All Versions]. [Project]. This paper presents an automatic evaluation framework for the task of planning experimental protocols, and introduces BioProt: a dataset of biology protocols with corresponding pseudocode representations.
-
From intention to implementation: automating biomedical research via LLMs - Science China Information Sciences, 2025. [All Versions]. Conventional biomedical research is increasingly labor-intensive due to the exponential growth of scientific literature and datasets. Artificial intelligence (AI), particularly large language models (LLMs), has the potential to revolutionize this process by automating various steps. Still, significant challenges remain, including the need for multidisciplinary expertise, logicality of experimental design, and performance measurements. This paper introduces BioResearcher, the first end-to-end automated system designed to streamline the entire biomedical research process involving dry lab experiments. BioResearcher employs a modular multi-agent architecture, integrating specialized agents for search, literature processing, experimental design, and programming. By decomposing complex tasks into logically related sub-tasks and utilizing a hierarchical learning approach, BioResearcher effectively addresses the challenges of multidisciplinary requirements and logical complexity. Furthermore, BioResearcher incorporates an LLM-based reviewer for in-process quality control and introduces novel evaluation metrics to assess the quality and automation of experimental protocols. BioResearcher successfully achieves an average execution success rate of 63.07% across eight previously unmet research objectives. The generated protocols, on average, outperform typical agent systems by 22.0% on five quality metrics. The system demonstrates significant potential to reduce researchers’ workloads and accelerate biomedical discoveries, paving the way for future innovations in automated research systems.
-
A human-machine interface for automatic exploration of chemical reaction networks - Nature Communications, 2024. [All Versions]. Autonomous reaction network exploration algorithms offer a systematic approach to explore mechanisms of complex chemical processes. However, the resulting reaction networks are so vast that an exploration of all potentially accessible intermediates is computationally too demanding. This paper introduces a STEERING WHEEL to guide an otherwise unbiased automated exploration. The STEERING WHEEL algorithm is intuitive, generally applicable, and enables one to focus on specific regions of an emerging network. It also allows for guiding automated data generation in the context of mechanism exploration, catalyst design, and other chemical optimization challenges.
-
Active learning accelerates electrolyte solvent screening for anode-free lithium metal batteries - Nature Communications, 2025. [All Versions]. Anode-free or ‘zero-excess’ lithium metal batteries offer high energy density compared to current lithium-ion batteries but require electrolyte innovation to extend cycle life. Due to the lack of universal design principles, electrolyte development for anode-free lithium metal batteries is slow and incremental and mainly driven by trial-and-error. This work demonstrates the use of active learning as an alternative approach to accelerate electrolyte discovery for anode-free lithium metal batteries. Unlike conventional data-intensive frequentist machine learning techniques, the active learning framework employs sequential Bayesian experimental design with Bayesian model averaging to efficiently identify optimal candidates in typical data-scarce and noisy label settings. Using capacity retention in real Cu||LiFePO4 cells as the target property, the approach integrates experimental feedback to iteratively refine predictions. Starting with just 58 data points from an in-house cycling dataset, the active learning framework explored a virtual search space of 1 million electrolytes, rapidly converging on optimal candidates. After seven active learning campaigns with about ten electrolytes tested in each, four distinct electrolyte solvents are identified that rival state-of-the-art electrolytes in performance. This work showcases the promise of active learning approaches in navigating large electrolyte chemical spaces for next-generation batteries.
-
PatCID: an open-access dataset of chemical structures in patent documents - Nature Communications, 2024. [All Versions]. The automatic analysis of patent publications has potential to accelerate research across various domains, including drug discovery and material science. Within patent documents, crucial information often resides in visual depictions of molecule structures. PatCID (Patent-extracted Chemical-structure Images database for Discovery) allows to access such information at scale. It enables users to search which molecules are displayed in which documents. PatCID contains 81M chemical-structure images and 14M unique chemical structures. This work compares PatCID with state-of-the-art chemical patent-databases. On a random set, PatCID retrieves 56.0% of molecules, which is higher than automatically-created databases, Google Patents (41.5%) and SureChEMBL (23.5%), as well as manually-created databases, Reaxys (53.5%) and SciFinder (49.5%). Leveraging state-of-the-art methods of document understanding, PatCID high-quality data outperforms currently available automatically-generated patent-databases. PatCID even competes with proprietary manually-created patent-databases. This enables promising applications for automatic literature review and learning-based molecular generation methods.
-
Large language models for scientific discovery in molecular property prediction - Nature Machine Intelligence, 2025. [All Versions]. Large language models (LLMs) are a form of artificial intelligence system encapsulating vast knowledge in the form of natural language. These systems are adept at numerous complex tasks including creative writing, storytelling, translation, question-answering, summarization and computer code generation. Although LLMs have seen initial applications in natural sciences, their potential for driving scientific discovery remains largely unexplored. This work introduces LLM4SD, a framework designed to harness LLMs for driving scientific discovery in molecular property prediction by synthesizing knowledge from literature and inferring knowledge from scientific data. LLMs synthesize knowledge by extracting established information from scientific literature, such as molecular weight being key to predicting solubility. For inference, LLMs identify patterns in molecular data, particularly in Simplified Molecular Input Line Entry System-encoded structures, such as halogen-containing molecules being more likely to cross the blood–brain barrier. This information is presented as interpretable knowledge, enabling the transformation of molecules into feature vectors. By using these features with interpretable models such as random forest, LLM4SD can outperform the current state of the art across a range of benchmark tasks for predicting molecular properties. The authors foresee it providing interpretable and potentially new insights, aiding scientific discovery in molecular property prediction.
-
Retrosynthesis prediction using an end-to-end graph generative architecture for molecular graph editing - Nature Communications, 2023. [All Versions]. Retrosynthesis planning, the process of identifying a set of available reactions to synthesize the target molecules, remains a major challenge in organic synthesis. Recently, computer-aided synthesis planning has gained renewed interest and various retrosynthesis prediction algorithms based on deep learning have been proposed. However, most existing methods are limited to the applicability and interpretability of model predictions, and further improvement of predictive accuracy to a more practical level is still required. In this work, inspired by the arrow-pushing formalism in chemical reaction mechanisms, the authors present an end-to-end architecture for retrosynthesis prediction called Graph2Edits. Specifically, Graph2Edits is based on graph neural network to predict the edits of the product graph in an auto-regressive manner, and sequentially generates transformation intermediates and final reactants according to the predicted edits sequence. This strategy combines the two-stage processes of semi-template-based methods into one-pot learning, improving the applicability in some complicated reactions, and also making its predictions more interpretable. Evaluated on the standard benchmark dataset USPTO-50k, the model achieves the state-of-the-art performance for semi-template-based retrosynthesis with a promising 55.1% top-1 accuracy.
-
ChipNeMo: Domain-Adapted LLMs for Chip Design - 2023. [All Versions]. ChipNeMo aims to explore the applications of large language models (LLMs) for industrial chip design. Instead of directly deploying off-the-shelf commercial or open-source LLMs, the authors instead adopt the following domain adaptation techniques: domain-adaptive tokenization, domain-adaptive continued pretraining, model alignment with domain-specific instructions, and domain-adapted retrieval models. The authors evaluate these methods on three selected LLM applications for chip design: an engineering assistant chatbot, EDA script generation, and bug summarization and analysis. Evaluations demonstrate that domain-adaptive pretraining of language models, can lead to superior performance in domain related downstream tasks compared to their base LLaMA2 counterparts, without degradations in generic capabilities. In particular, the largest model, ChipNeMo-70B, outperforms the highly capable GPT-4 on two of the use cases, namely engineering assistant chatbot and EDA scripts generation, while exhibiting competitive performance on bug summarization and analysis. These results underscore the potential of domain-specific customization for enhancing the effectiveness of large language models in specialized applications.
-
Single-atom alloy catalysts designed by first-principles calculations and artificial intelligence - Nature Communications, 2021. [All Versions]. This paper addresses the problem of new Single-atom-alloy catalysts (SAACs) discovery by applying a compressed-sensing data-analytics approach parameterized with density-functional inputs.
-
Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences - Proceedings of the National Academy of Sciences, 2021. [All Versions].
-
Comparability of automated human induced pluripotent stem cell culture: a pilot study - Bioprocess and Biosystems Engineering, 2016. [All Versions].
-
Virtual and augmented reality for biomedical applications - Cell Reports Medicine, 2021. [All Versions]. 3D visualization technologies such as virtual reality (VR), augmented reality (AR), and mixed reality (MR) have gained popularity in the recent decade. Digital extended reality (XR) technologies have been adopted in various domains ranging from entertainment to education because of their accessibility and affordability. XR modalities create an immersive experience, enabling 3D visualization of the content without a conventional 2D display constraint. This paper provides a perspective on XR in current biomedical applications and demonstrate case studies using cell biology concepts, multiplexed proteomics images, surgical data for heart operations, and cardiac 3D models. Emerging challenges associated with XR technologies in the context of adverse health effects and a cost comparison of distinct platforms are discussed. The presented XR platforms will be useful for biomedical education, medical training, surgical guidance, and molecular data visualization to enhance trainees’ and students’ learning, medical operation accuracy, and the comprehensibility of complex biological systems.
-
An augmented reality microscope with real-time artificial intelligence integration for cancer diagnosis - Nature Medicine, 2019. [All Versions]. The microscopic assessment of tissue samples is instrumental for the diagnosis and staging of cancer, and thus guides therapy. However, these assessments demonstrate considerable variability and many regions of the world lack access to trained pathologists. Though artificial intelligence (AI) promises to improve the access and quality of healthcare, the costs of image digitization in pathology and difficulties in deploying AI solutions remain as barriers to real-world use. This work proposes a cost-effective solution: the augmented reality microscope (ARM). The ARM overlays AI-based information onto the current view of the sample in real time, enabling seamless integration of AI into routine workflows.
-
Optimizing Spaced Repetition Schedule by Capturing the Dynamics of Memory - IEEE Transactions on Knowledge and Data Engineering, 2023. [All Versions].
-
LEGAL-BERT: The Muppets straight out of Law School - EMNLP'20, 2020. [All Versions]. Generating answers to legal questions, analyze contracts, and summarizing legal documents, making legal knowledge more accessible to non-experts.
-
BioBERT: a pre-trained biomedical language representation model for biomedical text mining - Bioinformatics, 2020. [All Versions]. Answering medical questions, identifying relevant clinical trials, and diagnosing diseases based on symptoms, making medical information more accessible to the general public.
-
Finbert: A pre-trained financial language representation model for financial text mining - IJCAI'20, 2020. [All Versions]. Predicting stock market trends, analyzing financial documents, and generating summaries of economic news articles, helping to disseminate financial knowledge.
-
SciBERT: A Pretrained Language Model for Scientific Text - EMNLP'19, 2019. [All Versions]. Searching and synthesizing scientific literature, aiding researchers in hypothesis generation, and assisting with experimental design, making scientific knowledge more accessible.
-
CodeBERT: A Pre-Trained Model for Programming and Natural Languages - EMNLP'20, 2020. [All Versions]. Completing code, generating programming documentation, and providing technical support, making programming knowledge more accessible to non-experts.
