Digital discovery最新文献

{"title":"optimade-maker: automated generation of interoperable materials APIs from static datasets","authors":"Kristjan Eimre, Matthew L. Evans, Bud Macaulay, Xing Wang, Jusong Yu, Nicola Marzari, Gian-Marco Rignanese and Giovanni Pizzi","doi":"10.1039/D6DD00125D","DOIUrl":"10.1039/D6DD00125D","url":null,"abstract":"<p >Atomistic structural data are central to materials science, condensed matter physics, and chemistry, and are increasingly digitised across diverse repositories and databases. Interoperable access to these heterogeneous data sources enables reusable clients and tools, and is essential for cross-database analyses and data-driven materials discovery. Toward this aim, the OPTIMADE (Open Databases Integration for Materials Design) specification defines a standard REST API for atomistic structures and related properties. However, deploying and maintaining compliant services remains technically demanding and poses a significant barrier for many data providers. Here, we present <em>optimade-maker</em>, a lightweight toolkit for the automated generation of OPTIMADE-compliant APIs directly from raw atomistic structure and property data. The toolkit supports a wide range of raw datasets, enables conversion to a standardised OPTIMADE data representation, and allows for rapid deployment of APIs in both local and production environments. We further demonstrate it through an automated service on the Materials Cloud Archive, which automatically creates and publishes OPTIMADE APIs for contributed datasets, enabling immediate discoverability and interoperability. In addition, we implement data transformation pipelines for the Cambridge Structural Database (CSD) and the Inorganic Crystal Structure Database (ICSD), enabling unified access to these curated resources through the OPTIMADE framework. By lowering the technical barriers to interoperable data publication, <em>optimade-maker</em> represents an important step toward a scalable, FAIR materials data ecosystem integrating both community-contributed and curated databases.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 6","pages":" 2469-2477"},"PeriodicalIF":7.1,"publicationDate":"2026-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13229571/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148159242","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Benchmarking explainable AI methods for toxicophore detection and toxicity prediction","authors":"Dina Khasanova and Igor V. Tetko","doi":"10.1039/D5DD00576K","DOIUrl":"https://doi.org/10.1039/D5DD00576K","url":null,"abstract":"<p >Recent studies have reported inconsistent behavior across explainable AI (XAI) methods in molecular property prediction, raising concerns about their reliability. This work investigates whether such inconsistencies arise from the XAI methods themselves or from the accuracy of the underlying predictive model. A high-accuracy model was first trained on deterministic functional-group labels, where all evaluated XAI methods consistently highlighted the correct atoms corresponding to the true structural motifs. The analysis was extended to mutagenicity prediction, where the methods again identified known toxicophores and chemically meaningful scaffolds. Model performance was then systematically degraded by introducing controlled amounts of label noise. As predictive accuracy decreased, agreement between XAI methods weakened gradually, and the highlighted features became less chemically relevant. When accuracy reached around 0.65, this trend changed, with a much sharper loss of agreement, indicating an explainability cliff. These findings underline the importance of assessing model accuracy before drawing conclusions from XAI outputs.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 6","pages":" 2547-2559"},"PeriodicalIF":7.1,"publicationDate":"2026-05-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://pubs.rsc.org/en/content/articlepdf/2026/dd/d5dd00576k?page=search","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148261441","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Constraint-aware labware layout generation from natural language for heterogeneous laboratory robots","authors":"Yuya Tahara-Arai, Takashi Inagaki, Akari Kato, Koji Ochiai, Kazuya Azumi, Koichi Takahashi, Genki N. Kanda and Haruka Ozaki","doi":"10.1039/D6DD00026F","DOIUrl":"https://doi.org/10.1039/D6DD00026F","url":null,"abstract":"<p >As laboratory automation accelerates scientific discovery, the manual, expertise-driven task of arranging labware on a robot's deck remains a significant barrier for researchers. This initial setup is crucial for experimental success, safety, and efficiency, yet its complexity grows exponentially with protocol intricacy. To address this challenge, we introduce Labware-Layout Planner, a novel system that integrates a Large Language Model (LLM)-based semantic interpreter with a spatial constraint solver. This architecture translates natural language experimental protocols into optimized, robot-agnostic labware layouts. By interpreting user instructions and physical constraints, our system automates the complex decision-making process of where to place each piece of equipment. We demonstrate its versatility through successful execution of diverse experiments: a liquid handling task and a complex qPCR assay on an Opentrons OT-2, and a multi-step cell passaging protocol on a humanoid Maholo LabDroid. Labware-Layout Planner represents a critical advance by tackling the physical setup phase of automation, paving the way for researchers to move seamlessly from a written idea to robotic execution and freeing human intellect for more creative scientific pursuits.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 6","pages":" 2507-2519"},"PeriodicalIF":7.1,"publicationDate":"2026-05-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://pubs.rsc.org/en/content/articlepdf/2026/dd/d6dd00026f?page=search","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148261439","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Molecular arms race classifier for decrypting venom peptide and ion channel interactions","authors":"Favour Achimba, Arezoo Bybordi, Mariam Gelashvili, Jessy Ramirez, Anita Raja, Weigang Qiu and Mandë Holford","doi":"10.1039/D5DD00498E","DOIUrl":"10.1039/D5DD00498E","url":null,"abstract":"<p >Animal venoms comprise an astonishing number of peptides, proteins and small molecules. The diversity of venom compounds arises from evolutionary adaptations resulting in both offensive and defensive traits in predators and prey alike. This concept, referred to as “arms race”, underpins the specificity and selectivity of venom compounds for certain molecular targets, like ion channels. Ion channels are essential regulators of cellular processes, and their dysfunction or dysregulation facilitates a wide range of diseases. Venom peptides and their derivatives are powerful modulators of ion channel activity, with several already in clinical use as FDA-approved therapeutics. Despite the remarkable potential of venom compounds, for the majority, their ion channel targets and modes of action remain largely uncharacterized. We hypothesize that venom peptides are constrained by and converge on molecular structures and targets despite their rapid sequence divergence due to arms race evolution. Here, we introduce a machine learning approach termed Molecular Arms Race Classifier (MARC), which predicts the ion channel targets – sodium, potassium, and calcium ion channels – of cysteine-rich venom peptides. MARC leverages evolutionary scale modeling (ESM) for feature extraction along with random forest classification to enable predictive functional annotation of venom compounds by their putative ion channel targets. MARC performs multi-class classification across four categories (sodium, calcium, potassium and non-ion-channels), to predict the ion channel targets of novel cysteine-rich venom compounds. We trained, tested, and cross-validated MARC on 5165 peptide and protein sequences sourced from in-house venom gland transcriptomes and public databases spanning diverse taxa, including sea anemones, snakes, scorpions, spiders, cone snails, and terebrid snails. We identified 28 novel terebrid snail venom peptides (teretoxins) predicted to target potassium ion channels. Orthogonal validation using docking and molecular dynamics simulations suggests the stability of the best docked pose of the K<small>⁺</small> channels, KcsA and MthK, to the teretoxin, Cje1.9, supporting MARC's robust prediction of Cje1.9 as a K<small>⁺</small> channel-targeting peptide. Taken together, these results indicate that MARC is a cost-effective method for screening vast peptide and protein libraries to identify potential ion channel targeting compounds, paving the way to design novel peptides selectively targeting distinct ion channel classes.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 6","pages":" 2575-2590"},"PeriodicalIF":7.1,"publicationDate":"2026-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13184802/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147984434","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessing the extrapolation capability of template-free retrosynthesis models","authors":"Jonghwi Choe, Shuan Chen and Yousung Jung","doi":"10.1039/D6DD00072J","DOIUrl":"https://doi.org/10.1039/D6DD00072J","url":null,"abstract":"<p >Template-free retrosynthesis models offer the potential to extrapolate beyond established chemical reaction spaces, addressing inherent limitations of template-based approaches. However, it remains unclear whether these models can reliably predict accurate, novel, and chemically feasible pathways outside their training distribution. In this study, we rigorously assess the extrapolation ability of state-of-the-art template-free models using carefully constructed out-of-distribution (OOD) benchmarks derived from USPTO datasets. While these models can generate novel synthetic routes, their exact-match accuracy on OOD reactions is remarkably low (typically &lt;1%). Moreover, round-trip performance (≈5–30%) is influenced by the performance of the forward model and may not fully capture some chemically reasonable predictions. Complementary manual inspection mitigates this limitation by revealing that the surrogate forward model produces false negatives, where chemically feasible reactions are incorrectly predicted as infeasible, and <em>vice versa</em> for false positives. These results underscore a critical challenge: current models may exhibit little creative extrapolation yet lack mechanisms to ensure chemical feasibility. Addressing this gap is essential for developing retrosynthesis models that are not only innovative, but also reliable for real-world synthesis planning.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 6","pages":" 2591-2599"},"PeriodicalIF":7.1,"publicationDate":"2026-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://pubs.rsc.org/en/content/articlepdf/2026/dd/d6dd00072j?page=search","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148261415","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Identification of drug candidates against glioblastoma with machine learning and high-throughput screening of heterogeneous cellular models","authors":"Vanessa Smer-Barreto, Richard J. R. Elliott, John C. Dawson, Álvaro Lorente-Macías, Muhammad Furqan, Asier Unciti-Broceta, Diego A. Oyarzún and Neil O. Carragher","doi":"10.1039/D5DD00190K","DOIUrl":"10.1039/D5DD00190K","url":null,"abstract":"<p >Glioblastoma multiforme (GBM) is an aggressive primary brain tumour that presents significant treatment challenges due to its complex pathology and heterogeneity. The lack of validated molecular targets is a major obstacle for discovering new therapeutic candidates, with no new effective GBM therapies delivered to patients in over two decades. Here, we report the identification of compounds that target the GBM stem cell survival phenotype. Our approach employs machine learning (ML) predictors of cell survival trained on high-throughput, image-based, phenotypic screening data for 3561 compounds, at multiple concentrations, across a panel of six heterogeneous, patient-derived, GBM stem cell lines. We computationally screened more than 12 000 compounds spanning various chemical classes. Experimental validation of ML-identified candidates across the GBM stem cell lines led to the identification of three compounds with activity against the GBM phenotype. Notably, one of our validated hits, the HSP90 inhibitor XL-888, displayed targeted elimination of all six GBM stem cell lines with IC<small>₅₀</small> in the nanomolar range. Further analyses suggest an XL-888 mechanism of action based on competitive ATP inhibition of HSP90 followed by disruption of HSP90 client proteins, and identify XL-888 as a promising candidate for future personalised medicine campaigns. Our work demonstrates that the use of phenotypic screening in tandem with ML can effectively identify therapeutic leads for personalised treatments in highly heterogeneous indications with few known molecular targets.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 6","pages":" 2560-2574"},"PeriodicalIF":7.1,"publicationDate":"2026-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13181845/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147977694","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CREOLab: a procedure captioning dataset for understanding creative tool use in object-rich laboratory videos","authors":"Shigeaki Goto and Tatsuki Hasebe","doi":"10.1039/D5DD00542F","DOIUrl":"https://doi.org/10.1039/D5DD00542F","url":null,"abstract":"<p >“Creative tool use” refers to the flexible application of tools beyond their intended purpose. In scientific experiments, this behavior is described as a “lab hack,” and its automatic documentation is valuable for accumulating experimental knowledge. Recently, vision-language models (VLMs) have shown promise for generating procedural descriptions from experimental videos. However, VLMs typically rely more on object-based knowledge than on understanding the manipulations. This issue is often overlooked in existing laboratory video datasets, as tools are typically used in standard, prescribed ways. Thus, the extent to which these models can interpret and describe actions that extend beyond object-based knowledge, such as creative tool use, remains uncertain. Moreover, laboratory environments often contain numerous items unrelated to the operation (<em>i.e.</em>, decoy objects), which can divert the model's attention and further complicate the accurate identification of creative manipulations. To address this limitation, we developed an evaluation dataset called “CREOLab” (CREative tool use in Object-rich Laboratories), consisting of 65 videos from 13 experimental scenarios featuring creative tool use, each recorded across five levels of decoy object density. Using a state-of-the-art, cloud-based VLM captioning system, we evaluated model performance. As the number of decoy objects increased, the model tended to insert redundant procedural steps or omit essential ones. As a result, it failed to document scenarios involving creative tool use accurately. These findings suggest that enhancing the reliability of automatic experimental recording with VLMs requires mechanisms for automated verification of generated outputs, as well as recording protocols that reduce the influence of decoy objects.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 6","pages":" 2600-2612"},"PeriodicalIF":7.1,"publicationDate":"2026-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://pubs.rsc.org/en/content/articlepdf/2026/dd/d5dd00542f?page=search","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148261452","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AI-designed and AI-implemented control systems for bespoke scientific instrumentation: application to scanning microscopy","authors":"Ke Wang, David J. Ward, Matthew Ord, Boyao Liu and Andrew P. Jardine","doi":"10.1039/D6DD00177G","DOIUrl":"https://doi.org/10.1039/D6DD00177G","url":null,"abstract":"<p >The pace of innovation in custom scientific instrumentation is frequently bottlenecked by the complexity of software engineering. While hardware designs evolve rapidly, developing robust and integrated control systems remains resource-intensive and often exceeds the software expertise available in experimental laboratories. Here, we present an AI-assisted workflow for constructing and validating an integrated control system for a bespoke scientific instrument, demonstrated on the Scanning Helium Microscope (SHeM). We emphasise that this work does not develop or train new AI models; instead, we use publicly accessible, general-purpose Large Language Models (LLMs) as practical engineering tools. Using these models, we co-develop a modular software stack spanning hardware interfaces, communication middleware, scan control, and user interfaces. To reduce the risk of deploying AI-generated code to fragile hardware, we introduced a digital sandbox that emulates instrument behaviour and supports pre-deployment verification, together with cross-validation using a second LLM and mandatory human review. We demonstrate successful deployment on a physical SHeM for 2D imaging and diffraction measurements, with behaviour consistent with a manually developed control system. This work provides a reproducible and safety-oriented template for AI-assisted software engineering in bespoke scientific instrumentation.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 6","pages":" 2448-2457"},"PeriodicalIF":7.1,"publicationDate":"2026-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://pubs.rsc.org/en/content/articlepdf/2026/dd/d6dd00177g?page=search","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148261463","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Vision-guided adaptive scooping for powder weighing in autonomous chemistry laboratories","authors":"Nikola Radulov, Thomas Little, Andrew I. Cooper and Gabriella Pizzuto","doi":"10.1039/D5DD00569H","DOIUrl":"https://doi.org/10.1039/D5DD00569H","url":null,"abstract":"<p >Autonomous, high-accuracy powder dispensing of heterogeneous solid materials remains an open challenge in automated chemistry laboratories. Existing systems often fail with diverse powder morphologies because they neglect the critical initial material acquisition step (<em>i.e.</em>, scooping). We present an end-to-end, vision-guided powder dispensing system that integrates an adaptive scooping mechanism with a deep reinforcement learning-based policy for dispensing. Our system utilises a parametrised scooping motion and assesses visually the acquired material volume after each scooping attempt. This visual feedback drives an iterative correction loop, allowing the robot to adjust its motion parameters to reject failures and obtain a suitable quantity for subsequent dispensing. We evaluated our system in real laboratory conditions using a set of 7 powders with varying physical properties. Our experiments demonstrate that the fully adaptive system outperformed fixed scooping baselines, achieving the lowest average absolute weighing error of 1.93 mg ± 2.04 mg across all materials.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 5","pages":" 2120-2127"},"PeriodicalIF":6.2,"publicationDate":"2026-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://pubs.rsc.org/en/content/articlepdf/2026/dd/d5dd00569h?page=search","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147968360","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Reaction center prediction by analyzing attention of a chemical language model","authors":"Xiaoliang Xiong, Ruizhen Jia, Yang Tian, Jingke Chen and Boxue Tian","doi":"10.1039/D6DD00055J","DOIUrl":"https://doi.org/10.1039/D6DD00055J","url":null,"abstract":"<p >Pretrained chemical language models are widely used to predict molecular properties and chemical reactions, yet interpreting their internal attention mechanisms remains difficult. Here, we analyze attention matrices from a chemical language model to extract information relevant to reaction center prediction. We introduce Peak-Activated Binary Attention (PABA), which binarizes attention matrices by retaining only peak values based on a parameter alpha. Using PABA, we identify a top-ranked attention head that effectively predicts reaction centers. We further develop a supervised extension, Supervised PABA (SPABA), which achieves a Matthews correlation coefficient (MCC) of 0.73 and outperforms existing supervised methods for reaction center prediction. SPABA reduces dependence on explicit reaction templates while preserving high accuracy and generalizability, providing a robust framework for reaction center prediction.</p>","PeriodicalId":72816,"journal":{"name":"Digital discovery","volume":" 6","pages":" 2458-2468"},"PeriodicalIF":7.1,"publicationDate":"2026-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://pubs.rsc.org/en/content/articlepdf/2026/dd/d6dd00055j?page=search","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148261464","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}