Xiangru Tang

Autonomous agents, from virtual laboratories to biological data-analysis agents and co-scientists, have begun to turn scientific goals into hypotheses. Whether they can go further and convert real, noisy and complex biological data into executable model hypotheses and architectures that encode testable assumptions remains unclear. Predicting how cells respond to genetic and chemical perturbations (i.e., modeling virtual cells) is a natural test of this ability: it is central to decoding disease mechanisms and therapeutic discovery, yet it demands cross-disciplinary expertise rarely concentrated in a single research group. Here, we introduce CellForge, an organized team of autonomous agents for single-cell perturbation modeling. Given a dataset and a task, it surveys the literature, parses the data, identifies relevant baselines, and formulates a model hypothesis through critic-mediated discussion among agents specialized in data modeling, single-cell biology, and deep learning. It then writes the training code, runs the experiments, and returns a trained predictor including architecture, code, and validation results. The architecture itself is the hypothesis: CellForge produces it end-to-end, with no human specifying the design, loss, or training protocol. Across six perturbation datasets spanning transcriptomic, epigenomic, and proteomic contexts, CellForge yields models that are competitive with hand-crafted state-of-the-art results on established tasks and extends the same workflow to paired multiomic readouts. Blinded human evaluators rate CellForge’s research plans as more scientifically sound and innovative than those from single LLMs or deep-research systems. In summary, we show autonomous agents can complete the full hypothesis-to-validation cycle in single-cell biology, with an AI effectively building another AI, in a complex setting that currently requires a team of human experts.

AI scientists powered by large language models have demonstrated substantial promise in autonomously conducting experiments and facilitating scientific discoveries across various disciplines. While their capabilities are promising, these agents also introduce novel vulnerabilities that require careful consideration for safety. However, there has been limited comprehensive exploration of these vulnerabilities. This perspective examines vulnerabilities in AI scientists, shedding light on potential risks associated with their misuse, and emphasizing the need for safety measures. We begin by providing an overview of the potential risks inherent to AI scientists, taking into account user intent, the specific scientific domain, and their potential impact on the external environment. Then, we explore the underlying causes of these vulnerabilities and provide a scoping review of the limited existing works. Based on our analysis, we propose a triadic framework involving human regulation, agent alignment, and an understanding of environmental feedback (agent regulation) to mitigate these identified risks. Furthermore, we highlight the limitations and challenges associated with safeguarding AI scientists and advocate for the development of improved models, robust benchmarks, and comprehensive regulations.

This survey addresses the crucial issue of factuality in Large Language Models (LLMs). As LLMs find applications across diverse domains, the reliability and accuracy of their outputs become vital. We define the “factuality issue” as the probability of LLMs to produce content inconsistent with established facts. We first delve into the implications of these inaccuracies. Subsequently, we analyze the mechanisms through which LLMs store and process facts, seeking the primary causes of factual errors. Our discussion then transitions to methodologies for evaluating LLM factuality, emphasizing key metrics, benchmarks, and studies. We further explore strategies for enhancing LLM factuality. Our survey offers a structured guide for researchers aiming to fortify the factual reliability of LLMs. We consistently maintain and update the related open-source materials at https://github.com/wangcunxiang/LLM-Factuality-Survey .

Inverse protein folding, which aims to design amino acid sequences for desired protein structures, is fundamental to protein engineering and therapeutic development. While recent deep-learning approaches have made remarkable progress, they typically represent biochemical properties as discrete features associated with individual residues. Here, we present BC-Design, a framework that represents biochemical properties as continuous distributions across protein surfaces and interiors. Through contrastive learning, our model learns to encode essential biochemical information within structure embeddings, enabling sequence prediction using only structural input during inference—maintaining compatibility with real-world applications while leveraging biochemical awareness. BC-Design achieves 88% sequence recovery versus state-of-the-art methods’ 67% (a 21% absolute improvement) and reduces perplexity from 2.4 to 1.5 (39.5% relative improvement) on the CATH 4.2 benchmark. Notably, our model exhibits robust generalization across diverse protein characteristics, performing consistently well on proteins of varying sizes (50-500 residues), structural complexity (measured by contact order), and all major CATH fold classes. Through ablation studies, we demonstrate the complementary contributions of structural and biochemical information to this performance. Overall, BC-Design establishes a new paradigm for integrating multimodal protein information, opening new avenues for computational protein engineering and drug discovery.

Large language models (LLMs) have dramatically enhanced the field of language intelligence, as demonstrably evidenced by their formidable empirical performance across a spectrum of complex reasoning tasks. Additionally, theoretical proofs have illuminated their emergent reasoning capabilities, providing a compelling showcase of their advanced cognitive abilities in linguistic contexts. Critical to their remarkable efficacy in handling complex reasoning tasks, LLMs leverage the intriguing chain-of-thought (CoT) reasoning techniques, obliging them to formulate intermediate steps en route to deriving an answer. The CoT reasoning approach has not only exhibited proficiency in amplifying reasoning performance but also in enhancing interpretability, controllability, and flexibility. In light of these merits, recent research endeavors have extended CoT reasoning methodologies to nurture the development of autonomous language agents, which adeptly adhere to language instructions and execute actions within varied environments. This survey article orchestrates a thorough discourse, penetrating vital research dimensions, encompassing (i) the foundational mechanics of CoT techniques, with a focus on elucidating the circumstances and justification behind its efficacy; (ii) the paradigm shift in CoT; and (iii) the burgeoning of language agents fortified by CoT approaches. Prospective research avenues envelop explorations into generalization, efficiency, customization, scaling, and safety. A repository for the related papers is available at https://github.com/Zoeyyao27/CoT-Igniting-Agent .

Chemical reasoning usually involves complex, multi-step processes that demand precise calculations, where even minor errors can lead to cascading failures. Furthermore, large language models (LLMs) encounter difficulties handling domain-specific formulas, executing reasoning steps accurately, and integrating code effectively when tackling chemical reasoning tasks. To address these challenges, we present ChemAgent, a novel framework designed to improve the performance of LLMs through a dynamic, self-updating library. This library is developed by decomposing chemical tasks into sub-tasks and compiling these sub-tasks into a structured collection that can be referenced for future queries. Then, when presented with a new problem, ChemAgent retrieves and refines pertinent information from the library, which we call memory, facilitating effective task decomposition and the generation of solutions. Our method designs three types of memory and a library-enhanced reasoning component, enabling LLMs to improve over time through experience. Experimental results on four chemical reasoning datasets from SciBench demonstrate that ChemAgent achieves performance gains of up to 46% (GPT-4), significantly outperforming existing methods. Our findings suggest substantial potential for future applications, including tasks such as drug discovery and materials science. Our code can be found at https://github.com/gersteinlab/chemagent

Inverse protein design (IPD) extends the classical problem of inverse protein folding (IPF) by not only recovering a sequence compatible with a given backbone geometry but also generating a range of new sequences that satisfy the physicochemical constraints required for a stable fold with a specified (often designed) function. Recent progress has incorporated biochemical properties as discrete per-residue features, but such localized signals cannot represent the continuous hydrophobic and electrostatic environments that span protein surfaces and interior volumes. Here we introduce BC-D esign , a biochemistry-aware inverse design framework that integrates geometric structure with smoothly varying hydrophobicity and charge properties. The later are represented independent of residue coordinates, on compact point clouds sampled throughout protein surfaces and interiors. This formulation provides a natural spatial description of physicochemical environments and enriches structural cues without specifying amino-acid identities. On the CATH 4.2 benchmark, BC-D esign achieves 90% sequence recovery and generalizes robustly across protein lengths, contact-order regimes, and major fold classes. Masking experiments further demonstrate an interpretable fidelity–diversity control: full biochemical context yields high-fidelity reconstructions, whereas withholding information enables exploratory variant generation. Beyond diagnostic benchmarks, BC-D esign demonstrably improves functional design outcomes across diverse settings. It increases enzyme–substrate affinity, enhances peptide–receptor design accuracy, and achieves state-of-the-art recovery and structural fidelity in antibody loop (CDRH3) modeling. In these applications, the model not only reconstructs native-like sequences but also proposes plausible functional variants that preserve catalytic or binding geometry—such as generating CDRH3 loops that maintain antigen-contact configurations while offering new sequence solutions. These results show that integrating continuous physicochemical properties with structural geometry enables practical, function-oriented protein design, going beyond backbone-conditioned recovery.

Motivation: The current paradigm of deep learning models for the joint representation of molecules and text primarily relies on 1D or 2D molecular formats, neglecting significant 3D structural information that offers valuable physical insight. This narrow focus inhibits the models’ versatility and adaptability across a wide range of modalities. Conversely, the limited research focusing on explicit 3D representation tends to overlook textual data within the biomedical domain. Results: We present a unified pre-trained language model, MolLM, that concurrently captures 2D and 3D molecular information alongside biomedical text. MolLM consists of a text Transformer encoder and a molecular Transformer encoder, designed to encode both 2D and 3D molecular structures. To support MolLM’s self-supervised pre-training, we constructed 160K molecule-text pairings. Employing contrastive learning as a supervisory signal for learning, MolLM demonstrates robust molecular representation capabilities across four downstream tasks, including cross-modal molecule and text matching, property prediction, captioning, and text-prompted molecular editing. Through ablation, we demonstrate that the inclusion of explicit 3D representations improves performance in these downstream tasks. Availability and implementation: Our code, data, pre-trained model weights, and examples of using our model are all available at https://github.com/gersteinlab/MolLM. In particular, we provide Jupyter Notebooks offering step-by-step guidance on how to use MolLM to extract embeddings for both molecules and text.