Biology subjects

Yang, Q.

Publications and source records attributed to Yang, Q..

At least 19 recordsLinked to original sources

bioRxiv2026

A polysynaptic hypothalamuspelvic-ganglion circuit regulates prostate tumor growth via WNT3A

Brain-tumor communication is increasingly recognized. Peripheral autonomic nerves regulate prostate cancer progression, yet how the brain influences local nerve-tumor interactions remains unclear. We identify a polysynaptic circuit linking the hypothalamic paraventricular nucleus (PVN), dorsal motor nucleus of the vagus (DMV), major pelvic ganglion (MPG) and prostate tumor in mice. Although both nuclei were activated during tumor growth, PVN activity restrained, whereas DMV activity promoted tumor growth. MPG ablation abolished DMV-induced tumor promotion. Combining LMD with Smart-seq2 revealed transcriptional remodeling and increased Wnt3a expression in tumor-innervating, activated MPG neurons under upstream neural control. MPG explants released WNT3A, whose neutralization suppressed explant-induced cancer-cell proliferation. Together, our findings identify a polysynaptic PVN-DMV-MPG circuit that regulates prostate tumor growth and implicate WNT3A as a mediator of the pro-tumor effects of the MPG, highlighting the therapeutic potential of targeting the neural circuit and downstream molecular pathway.

cancer biology↗

bioRxiv2026

ANO6 Scaffolds ER Mitochondria Contacts to Govern Spatial Lipid Compartmentalization and Ferroptosis Evasion in Chemoresistant Breast Cancer

Background: Metabolic glycerophospholipid remodeling promotes tumor persistence under neoadjuvant chemotherapy in breast cancer; however, how resistant cells evade lipotoxicity from accumulating peroxidizable lipids presents an intriguing metabolic conundrum. The underlying organelle-level spatial mechanisms remain poorly defined. Methods: Integrated multi-omic profiling of paired pre- and post-NAC clinical specimens was performed. Engineered ANO6 models (wild-type, knockout, and conformational mutants) were characterized using subcellular lipidomics of purified mitochondria-associated ER membranes (MAMs), electron microscopy, in situ proximity ligation assays, and preclinical xenograft models. Results: Chemoresistant lesions exhibited marked ANO6 upregulation and polyunsaturated phospholipid (PUFA-PL) accumulation without enhanced apoptosis. Beyond its canonical role as a plasma membrane scramblase, ANO6 operates as an organelle contact scaffold at MAMs, leveraging its conformational activation to coordinate inter-organellar tethering. ANO6 directly recruits vimentin (VIM) and VAPB to assemble a VIM-ANO6-VAPB-PTPIP51 tethering platform, structurally bridging intermediate filaments to organelle contact sites. This complex tightens ER-mitochondria coupling, establishing a specialized structural platform that routes and sequesters peroxidizable lipids into protective cytoplasmic lipid droplets, thereby shielding functional membrane bilayers from ferroptotic peroxidation. Functional dissection revealed that the active scaffolding conformation (I612A) sustains MAMs integrity and lipid packaging, whereas scaffold deficiency (D703R) compromises contact architecture. Pharmacologically perturbing this platform with niclosamide destabilized lipid sequestration, selectively resensitizing chemoresistant cells to synthetic lethal ferroptosis in vitro and in vivo. Conclusions: ANO6-mediated contact scaffolding governs spatial lipid compartmentalization, establishing the ANO6-MAMs structural axis as a potential target to exploit lipid metabolic vulnerabilities in refractory breast cancer.

cancer biology↗

bioRxiv2026

DCSMAP: A Hybrid Linear-Pangenome Framework for Accurate and Efficient Read Alignment

Background: Linear-reference read alignment is computationally efficient but is susceptible to reference bias, whereas pangenome graph alignment better represents genetic diversity at substantially greater computational cost. Findings: We developed DCSMAP, a hybrid workflow that combines linear alignment with machine-learning-based selection of reads for remapping to a personalized pangenome graph, followed by local indel realignment. An XGBoost classifier trained on linear-alignment features achieved an area under the receiver operating characteristic curve of 0.9831 on held-out test data. Across 17 runs covering seven Genome in a Bottle samples, two sequencing platforms, and two read lengths, it selected a mean of 14.80% of read pairs for graph remapping. The optimized hybrid workflows reduced variant-calling errors by 46.66% relative to BWA-MEM2 and 8.39% relative to vg personalized-pangenome alignment. Heterozygous allelic balance indicated that the hybrid workflows substantially reduced reference bias relative to BWA-MEM2 linear alignment, yielding reference bias comparable to that of the pangenome alignment. On HG003 DNBSEQ PE150 data, DCSMAP achieved a 2 to 2.5 times speedup over the personalized-pangenome workflow; both workflows used 32 threads. Efficiency gains arose from both the hybrid architecture and tool-level optimizations, which accelerated personalized-graph construction by 2.5 times, graph alignment by 1.2 times, and indel realignment by 2.5 times. Conclusions: Selectively applying graph alignment to reads with potentially unreliable linear placements retains the accuracy benefits of personalized pangenomes while substantially reducing computation. DCSMAP provides an accurate and resource-efficient approach to short-read mapping on a standard CPU server.

bioinformatics↗

bioRxiv2026

DeepD enables efficient genome modeling and controllable long-context sequence generation across domains of life

Genome foundation models hold great potential for advancing biological research, but modeling diverse complex genomes and generating controllable genome-scale sequences remain challenging. We present DeepD, a phylogeny-informed sparse mixture-of-experts model operating at single-nucleotide resolution over a 1-million-token context while reducing active computation. Pretrained on unselected whole-genome sequences across all domains of life, DeepD learns genic and intergenic sequence features and supports broad sequence-to-function prediction across species. DeepD performs well in human protein fitness and disease variants effect prediction, such as BRCA1 and BRCA2, with particular strength in regulatory regions, such as promoter and transcription-factor binding sites. Expert routing captures exon-intron boundaries, transposable elements, and distal cis-regulatory elements. DeepD also generates experimentally validated core promoters with specified transcriptional strengths. Finally, the Context-Anchor-Body strategy enables generation of complete yeast chromosome sequences and megabase-scale Arabidopsis segments, preserving gene structure and order while introducing variations within natural ranges. Together, DeepD is an efficient genome foundation model that unifies accurate sequence-to-function prediction with controllable genome-scale generation, advancing predictive and generative design of complex genomes.

bioinformatics↗

bioRxiv2026

Dynamics of dopamine transients in human olfactory epithelium during reward learning

Dopamine signaling is crucial for motivation, movement, and learning, but has remained beyond the reach of measurement in healthy humans. Here, we measured sub-second dopamine in the olfactory epithelium of healthy humans during two reward learning tasks with distinct reward types: liquid reward during a conditioning task and monetary reward during an economic exchange game. In both tasks, as trials progress, peak dopamine transients shift from rewarding outcomes back to reward-predictive cues. In the economic game, this pattern retrains quickly when reward magnitude or reward timing is changed without warning. These findings match the time transfer characteristics and amplitude dependence of temporal difference prediction error signals. The results also establish a new, minimally invasive route for recording fast dopamine signaling in healthy humans: the nose.

neuroscience↗

bioRxiv2026

Impacts of different SETD1A mutations and genetic backgrounds on human induced pluripotent stem cell-derived hippocampal neurons

Abstract Human induced pluripotent stem cells (hiPSCs) provide a translational platform to interrogate the impact of genetic risk factors for various disorders on human biology. It has been a common practice to engineer premature stop codons to model predicted loss-of-function mutations using isogenic hiPSCs, yet effects of mutation position for the same gene and genetic background have been rarely tested systematically and at multiple levels for comparison to those of patient-specific mutations. Ultra-rare, heterozygous, and predicted loss-of-function mutations in the SETD1A gene, which encodes a H3K4 methyltransferase, are highly penetrant genetic risk factors for schizophrenia. Here, we generated a large panel of isogenic heterozygous SETD1A mutant hiPSC lines, including premature stop-codon mutations in one early exon and one late exon in five genetic backgrounds and two patient-specific mutations, to gain insight into molecular, cellular, and functional changes in a disease-relevant and highly pure population of human hippocampal dentate granule cell-like neurons. Our multi-level analyses showed that different SETD1A mutations lead to divergent molecular and cellular phenotypes related to synapses, but convergent neuronal network deficits, while the genetic background influences gene expression with no substantial effects at cellular and network levels. Mechanistic analysis confirmed the early exon patient mutation as loss-of-function, but revealed that the predicted loss-of-function patient mutation in the late exon is likely gain-of-function. Consequently, while all mutations decreased global H3K4me3 levels, deficits were only partially rescued by inhibiting H3K4 demethylase LSD1 for early exon mutations, but not for late exon mutations. Together, these results suggest that highly penetrant mutations can lead to similar cellular and functional phenotypes across different genetic backgrounds, whereas different premature stop-codon mutations could have contrasting impacts due to distinct mechanisms. Our findings have practical implications for applying genome-editing in hiPSCs to model disorders, highlight the value of modeling patient-specific mutations and testing multiple mutations, and suggest the need for tailored mutation-dependent therapeutic strategies.

neuroscience↗

bioRxiv2026

Targeting ribosome biogenesis sensitizes PARPi-resistant triple-negative breast cancer

Poly (ADP-ribose) polymerase inhibitors (PARPi) are the only targeted treatment available for early triple-negative breast cancer (TNBC), an aggressive subtype with limited treatment options. While PARPi are effective in patients with BRCA mutations, resistance to these inhibitors can arise. We previously developed BRCA-mutant cell lines, HCC1395, with acquired resistance to talazoparib and olaparib, and have now developed PARPi-resistant MX1 cell lines in an analogous manner. Ribosomal pathways were upregulated in 3/4 cell lines with acquired PARPi resistance. The resistant cell lines showed high and low basal nucleolar stress based on nucleolar morphology and nucleophosmin delocalization from nucleolus to nucleoplasm. Treatment of these cell lines with the combination of olaparib and a ribosome biogenesis inhibitor, BMH-21, demonstrated synergy in the PARPi-resistant cell lines, but not in the parental cells. Interestingly, HCC1395 cells, which had lower baseline nucleolar stress, showed greater nucleophosmin delocalization with combination treatment than MX1 cells, which had higher baseline nucleolar stress. All four cell lines demonstrated sustained DNA damage with the combination. We extended our investigation to 3 BRCA wild-type TNBC cell lines, BT549, Hs578T, and MDAMB231 and identified synergy in the context of intrinsic PARPi resistance. In MDAMB231 cells, combining olaparib with BMH-21 induced a senescence-like phenotype, increasing gH2AX-positive cells by up to 7.3-fold, senescence-associated B-galactosidase (SA-B-gal) staining by 5.6-fold, and p21 by 3.6-fold. These results were mirrored with a second ribosome biogenesis inhibitor, CX-5461. Finally, we identified distinct patterns of ribosomal pathway activity in a cohort of patients with treatment-resistant TNBC, and found correlations between ribosomal pathways with MYC, proliferation, DNA damage, and senescence. Together, these data identify ribosome biogenesis as a potential therapeutic vulnerability in PARPi-resistant TNBC and support its targeting in combination with PARPi, while demonstrating a broader for ribosome biogenesis in treatment-resistant TNBC.

cancer biology↗

bioRxiv2026

Nutrient-sensitive histone propionylation regulates T cell fate and function

Tissue microenvironments differ widely in nutrient composition; yet how local nutrient conditions shape epigenetic programs that direct immune cell fate is poorly understood. We identify histone lysine propionylation (Kpr) as a chromatin mechanism linking nutrient availability to T cell fate decisions. Proteomic analysis reveals that Kpr is broadly reduced in exhausted T cells. H3K18pr is enriched at stemness-associated loci in effector T cells and depleted upon exhaustion. Isoleucine and propionate supply propionyl-CoA in T cells, and restricted nutrient availability within the tumor microenvironment limits histone Kpr, promoting T cell exhaustion. Disrupting isoleucine catabolism in T cells further suppresses Kpr, without altering histone acetylation, and impairs T cell function and antitumor immunity in a manner rescued by propionate. Dietary supplementation of isoleucine and propionate increases Kpr in tumor infiltrating lymphocytes, limits exhaustion, and improves tumor control. These findings establish a microenvironment-chromatin axis crucial for antitumor immunity and amenable to nutritional intervention.

immunology↗

bioRxiv2026

Unified modeling of mutation-induced protein stability responses with OmegaTherm

Accurate prediction of mutation-induced protein stability responses, including changes in folding free energy and melting temperature, is critical for protein engineering. Existing models represent these related but non-equivalent measurements separately and often focus on single mutations. Here we present OmegaTherm, a sequence-based framework that jointly learns both stability responses for single and multiple mutations. OmegaTherm introduces a shared protein language model to capture transferable mutation context, measurement-specific encoders to preserve measurement identity, and paired-measurement-informed distillation to learn complementary knowledge from heterogeneous data. An antisymmetric prediction architecture further enforces physical consistency between forward and reverse mutations. Across independent benchmarks, OmegaTherm achieves state-of-the-art performance in quantitative prediction, mutation classification and protein-level ranking. Beyond benchmark prediction, OmegaTherm-derived stability landscapes correlate strongly with fitness landscapes from deep mutational scanning, enrich favorable variants and reveal function-associated mutation-sensitive regions. These results establish unified representation learning as an effective strategy for integrating related but non-equivalent biochemical measurements.

bioinformatics↗

bioRxiv2026

Predictive Modeling of Cancer Cell Growth Kinetics with Machine Learning

Quantitative characterization of cell proliferation is central to preclinical drug discovery. Here, we evaluated Random Forest (RF) regression for predicting confluence-based cell growth trends using data from human cancer cell lines and benchmarked its performance against the widely used logistic and Gompertz models. The RF model achieved higher predictive accuracy within the evaluated dataset, particularly at higher initial seeding densities. We further extended the RF framework to predict growth curves across seeding densities. Collectively, these findings highlight the potential of data-driven modeling to complement conventional mathematical approaches and support more informed cell-culture design.

bioinformatics↗

bioRxiv2026

The transcription factor THAP11 is a general regulator of early embryonic polarization

During mouse embryogenesis, three sequential waves of cell polarization occur in the 8-cell-morula trophectoderm (TE), E3.5-E4.5 primitive endoderm (PrE), and E4.5-E5.5 epiblast (Epi), driving cell fate determination and morphogenic events that lay the foundation for embryonic development. However, it is not clear whether a general regulator that functions in all three waves of polarization exist. Here, we demonstrate that transcriptional factor THAP11 is such a general regulator. Using a novel mouse model combining genetic knockout and an inducible degron system, we demonstrate that stage-specific acute THAP11 depletion causes the failure of the three waves cell polarization, resulting in defective blastocyst formation, PrE maturation, and Epi lumenogenesis, respectively. Mechanistically, THAP11 does not directly regulate pluripotency genes but drive polarization through both stage-specific and conserved chromatin-bindings that commonly activate polarity programs, including Golgi vesicle transport. THAP11 activates polarity genes by promoting chromatin accessibility at their promoters. THAP11, together with stage-specific regulators, sequentially paces the timing of cell polarization. Collectively, our study establishes THAP11 as a general regulator of embryonic morphogenesis.

developmental biology↗

bioRxiv2026

Maleic acid adjuvates BCG to induce CD4+CD8+ T cells and sterilize tuberculosis

Anti-tuberculosis (TB) immunity typically emphasizes CD4+ T and CD8+ T cells that only provide limited protection against Mycobacterium tuberculosis (Mtb) infection. The anti-TB function of CD4+CD8+ T cells remains unexplored, mostly due to the dogma that ThPOK and Runx3 antagonize each other to block their development and their scarcity. Here, we show that immunization of BCG (Bacillus Calmette-Gu&eacuterin) with Complete Freund's Adjuvant (CFA) markedly induced CD4+CD8+ T cells through the gut microbiota Clostridium butyricum or its metabolite maleic acid. Mtb-specific CD4+CD8+ T cells exhibited a broad T-cell receptor repertoire and sterilized Mtb infection by inducing SLAMF1 and neutrophil responses. Mechanistically, maleic acid interacted with PAK2 to promote STAT5 phosphorylation at Ser127, thus upregulating Ets1 expression. Direct binding of Ets1 to the Runx3 and ThPOK promoters orchestrated their expression to redirect the CD8+ T cell differentiation towards the CD4+CD8+ lineage. Moreover, immunization of BCG plus Clostridium butyricum or maleic acid sterilized Mtb infection in mouse models. These findings establish CD4+CD8+ T cells as superior immune effectors against Mtb, a concept that could fundamentally redirect the current paradigm for anti-TB vaccine development.

microbiology↗

bioRxiv2026

Multi-Omics analysis provides crucial insights into ecological adaptation to dryland of a dominant grass (Psammochloa villosa, Poaceae) in Northwest China

Desertification exerts dramatic selection pressures on the evolution of plants. Despite the key role of ecological adaptation by natural selection to arid grasslands and subsequent intraspecific divergence, specific mechanisms driving this process remain poorly understood. Psammochloa villosa, a perennial forage grass endemic to the arid grasslands in Northwest China, where it thrives in shifting and semi-fixed sand land due to its exceptional drought tolerance, provides an ideal system to study adaptive evolution to aridity. In our study, we assembled a high-quality, chromosome-scale genome and conducted genomic resequencing of 42 populations across its major distribution. The genome assembly, which is approximately 1.55 Gb in size, has a super-scaffold N50 of 66.79 Mb, with 75.84% of the sequences identified as transposable elements. Coalescent phylogeny and genomic collinearity analyses strongly supported that P. villosa and Neotrinia splendens, as the closest taxa, shared a recent whole-genome duplication (WGD) event occurring approximately 18-20 Mya and followed by their divergence around ~11.2 Mya. Based on ancestral grass karyotype (AGK) reconstruction from synteny analysis, our results suggest that, relative to the AGK after the {rho}-WGD event, P. villosa and N. splendens underwent similar chromosomal restructuring and lineage-specific retention of numerous copies, providing a genomic basis for potential ecological adaptation and intraspecific diversification. The expanded XTH family, which encodes enzymes mediating xyloglucan endotransglucosylation and hydrolysis and thereby regulating xyloglucan remodeling, showed strong transcriptional responses under PEG-6000 treatment, suggesting that retained copies may be associated with xerophytic adaptation in P. villosa. Together, these findings suggest that WGD-derived gene retention created a delayed reservoir of genetic diversity that was later shaped by desertification and Qinghai-Xizang Plateau environmental changes, contributing to climate-associated genomic islands and intraspecific differentiation.

genomics↗

bioRxiv2026

Avs2 drives non-canonical tetrameric assembly of trypsin-like domain for anti-phage defense

Bacteria have evolved diverse anti-phage defense systems, with the antiviral STAND family (Avs) representing one of the most diverse and widespread, encompassing at least 90 distinct families, yet only Avs3, Avs4, Avs5 and Avs7 have been well characterized. Here, we elucidated the molecular mechanism of Avs2-trypsin-MBL system, where phage terminase recognition by Avs2 triggers coupled activation of the protease and nuclease activities of trypsin-MBL. Cryo-EM structure of Avs2-trypsin-terminase and biochemical analysis reveal that the binding of terminase ATPase domain triggers the assembly of Avs2 into tetramer, with two unique ATP molecules bridging ATPase active-site recognition by TPR domain. This tetramerization drives the fused trypsin into an active C4-symmetric assembly, an architecture distinct from the conventional non-defense trypsin. Our study unravels the activation mechanism of Avs2-trypsin-MBL system, expanding our understanding on commonality and diversity of widespread Avs-mediated anti-phage immunity, alongside the structural and functional adaption of trypsin.

microbiology↗

bioRxiv2026

Context-dependent variant interpretation from Mendelian disease to genetic predisposition: a proof-of-concept using LPL

As genomic sequencing evolves beyond rare disease diagnostics toward population screening and precision medicine, clinical variant interpretation is increasingly challenged by variants whose clinical consequences depend on biological context. Current frameworks, including the ACMG/AMP guidelines, generally assign a single classification to each variant regardless of inheritance state or genetic context, potentially failing to communicate context-dependent clinical consequences. Here, we address this issue using loss-of-function variants in LPL as a uniquely informative model system in which residual physiological LPL activity can be directly quantified in vivo. By systematically integrating published biallelic LPL genotypes, physiological measurements, functional studies, and clinical phenotypes, we identified a biologically meaningful transition at approximately 10% residual physiological LPL activity. Activity below this level was predominantly associated with classical childhood-onset familial chylomicronemia syndrome (FCS), whereas higher activity was associated with phenotypic attenuation and modifier-dependent clinical expression. Furthermore, heterozygous loss-of-function variants exhibited an estimated penetrance of 5-7% for severe hypertriglyceridemia. We therefore propose a context-dependent framework in which biallelic complete- or near-complete loss-of-function genotypes are interpreted as causative for FCS, whereas heterozygous variants are interpreted as predisposing to severe hypertriglyceridemia while retaining recognition of FCS carrier status. Together, our findings demonstrate that clinical variant interpretation should integrate available biological context--including, where relevant, allelic configuration, residual biological function, and penetrance--rather than rely on the intrinsic molecular consequence of the variant alone. More broadly, this framework provides a conceptual model for interpreting variants across the continuum from Mendelian disease to genetic predisposition in the era of precision medicine.

genetics↗

bioRxiv2026

BCIJelly: An integrated ecosystem for brain-computer interface research

Brain-computer interface (BCI) research relies on multistage computational pipelines, but progress has been slowed by fragmented data formats, heterogeneous decoder implementations and hardware-specific deployment toolchains. Here, we introduce BCIJelly, a unified ecosystem that standardizes 18 BCI datasets into AI-ready inputs and integrates 15 benchmark decoders, 80 reusable modules, automated architecture search (AAS) and hardware-aware neuromorphic deployment. Our AAS constructs task-specific decoders without manual design and extends into a large language model (LLM)-driven closed-loop mode supporting single-task, multitask and cross-species decoder design. A single-command pipeline compiles trained decoders for neuromorphic hardware, reducing power consumption by 30 to 50 times while preserving decoding performance. An interactive visualization software enables code-free exploration of neural recordings and decoding outputs. BCIJelly is validated across five BCI paradigms (motor, visual, speech, emotion and auditory) in humans, macaques and mice, providing an extensible ecosystem connecting data standardization, decoder development, systematic evaluation and hardware-aware deployment for BCI research.

neuroscience↗

bioRxiv2026

Spatial cholesterol homeostasis gatekeeps T-cell development and activation by orchestrating signaling and fitness

Cholesterol is essential for T-cell immunity, and its spatial distribution is tightly regulated. Although cholesterol is synthesized in the endoplasmic reticulum (ER), it is predominantly transported to the plasma membrane (PM); however, the machinery mediating this anterograde transport in T cells remains unknown. Here, through a functional genetic screen, we identify oxysterol-binding protein (OSBP) as the principal mediator of ER-to-PM cholesterol transport in T cells. OSBP deficiency depletes accessible PM cholesterol while causing cholesterol accumulation in the ER, resulting in impaired T-cell receptor (TCR) signaling and disruption of ER homeostasis. Using stage-specific conditional knockout mice, we demonstrate that OSBP is required at multiple developmental checkpoints in the thymus, including {beta}-selection, positive selection, and post-selection maturation. Loss of OSBP during early thymocyte development causes a near-complete block in T-cell development, resulting in a profound absence of mature peripheral T cells. In mature T cells, activation markedly increases dependence on OSBP-mediated cholesterol transport, with its inhibition causing ER perturbation and extensive cell death. Finally, we show that disease-associated oxysterols disrupt OSBP-mediated cholesterol transport, leading to T-cell dysfunction and providing a mechanistic explanation for impaired T-cell immunity in pathological settings. Together, our findings identify OSBP as a central regulator of intracellular cholesterol transport that couples membrane cholesterol homeostasis to TCR signaling and ER integrity. These results establish the spatial distribution of cholesterol, rather than its abundance alone, as a fundamental metabolic determinant of thymocyte development and peripheral T-cell function.

immunology↗

bioRxiv2026

An AI System for Autonomous Algorithm Evolution in Drug Development

Artificial intelligence (AI) is increasingly permeating the drug development pipeline. Numerous algorithms for accelerating this multi-stage and multi-task process have been constructed, which depends heavily on expert design and labor-intensive task-specific optimization. Given that AI-driven acceleration of drug development is recognized as a cumulative, often synergistic, effect across multiple stages, the autonomous evolution of existing algorithms across the entire pipeline is demanded to achieve a holistic advancement. Here, we present DrugEvolve, a multi-role large language model system for systematic and autonomous algorithm evolution in drug development. DrugEvolve realizes a closed-loop evolution process by incorporating Researcher, Engineer, and Analyst domains, and enables an iterative design, implementation, evaluation, and refinement of algorithm by leveraging scientific knowledge and accumulated evolutionary experience. Across eleven representative tasks spanning target identification, drug discovery, preclinical study, and clinical trial, DrugEvolve autonomously evolved the corresponding task-specific algorithms and achieved substantial performance enhancement on 120 benchmark test sets. Moreover, it showed robust generalizabilities across heterogeneous data modalities (ranging from biological sequence and graph to molecular topology and textual language), and realized gains in both predictive and generative tasks. Collectively, this AI system can serve not only as an algorithmic infrastructure for drug development, but also as a transferable paradigm for broader scientific domains.

pharmacology and toxicology↗