Biology subjects

Mi, T.

Publications and source records attributed to Mi, T..

10 recordsLinked to original sources

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↗

bioRxiv2026

Jump-starting the T cell response in established tumors

Checkpoint blockade only works in 10-20% of patients. Consequently, investigators are testing checkpoint inhibitors in combination with drugs like the class I histone deacetylase inhibitor, entinostat (ENT). Unfortunately, the combination of ENT and checkpoint blockade fared poorly in patients with breast or ovarian cancer, despite promising pre-clinical results. Here we show that ENT enhances CD8+ T cell responses by maintaining a progenitor-like population of CD8+ T cells that supplies activated effector T cells to tumors for prolonged periods. Surprisingly, the anti-tumor effects of ENT are only experienced when delivered during a narrow window that occurs after T cell activation and before T cell exhaustion--a window that is likely closed in most patients. However, by first "jump-starting" the T cell response using an oncolytic virus, the anti-tumor activity of ENT and PD1 blockade is restored. These data establish a general paradigm, independent of tumor type, to rationally manipulate anti-tumor immunity.

immunology↗

bioRxiv2026

Reprogramming CAR T-Cells with designed bioPROTACs

Gene editing has been used to enhance CAR T-cell function by disrupting negative regulators but has limitations. Here we show that de novo-designed generated targeted degraders (bioPROTACs) provide an alternative approach. Expression of bioPROTACs in CAR T-cells targeting DNMT3A, a key regulator of T-cell exhaustion, phenocopied gene knockout. Our reversible, non-gene editing approach provides a tunable strategy to reprogram T-cell fate which should be broadly applicable for next-generation cell therapies.

immunology↗

bioRxiv2025

Structural and Functional Insights into Alkaline-pH Sensing by the Alka Channel

The ability to detect environmental pH is essential for survival. In Drosophila melanogaster, the taste receptor Alka mediates alkaline food sensing, but the structural and functional mechanisms of Alka-mediated high-pH sensation have remained unclear. Here, we report the cryo-electron microscopy structure of Alka at near-atomic resolution. Alka assembles into a homopentamer and undergoes conformational changes between neutral and alkaline pH conditions. At nearly neutral conditions, a lysine residue (K229) in the channel activation domain forms a stabilizing salt bridge, maintaining the channel in a closed state. Under alkaline conditions, deprotonation of the K229 side chain disrupts this interaction, resulting in channel opening. Functional studies in heterologous cells and intact flies show that mutation of K229 significantly impairs alkaline pH-dependent Alka activation, demonstrating its essential role in pH sensing. Our findings uncover a previously unrecognized structural mechanism underlying alkaline pH detection and identify a lysine residue as a molecular switch for alkaline pH-mediated channel gating. Moreover, our work establishes hydroxide ions as a distinct class of ion channel ligands and reveals a molecular innovation of chemosensory receptors by which animals broaden their range of chemical senses to detect and avoid strongly alkaline environments.

biophysics↗

bioRxiv2025

Accelerating Virtual Directed Evolution of Proteins via Reinforcement Learning

With the advancement of machine learning methods, the protein fitness landscape can be predicted, providing reliable guidance in the selection of advantageous mutations for the directed evolution of proteins. However, the potential multiple mutational variants derived from the simple combi-nation of a limited number of advantageous single mutations may not represent superior choices. Moreover, the exploration and selection of the astronomical number of multiple mutational variants remain a highly challenging task. In this study, we introduce a virtual directed evolution pipeline, RelaVDEP, for the rapid identification of mutational variants with explicit enhancement in the de-sired property of the target protein. By adapting and fine-tuning a pre-trained fitness predictor to improve sequence-based protein functional prediction and by designing a model-based reinforce-ment learning framework to efficiently explore the vast combinatorial space of protein mutations, this pipeline is capable of effectively accelerating the directed evolution process for a broad spec-trum of proteins with versatile functional profiles. According to a series of experimental validations, the diversified mutational variants identified by our method exhibit notable improvements in desir-able protein functional properties. In particular, by integrating RelaVDEP with active learning, we successfully optimized the sequence of a PETase, enhancing its catalytic activity through previously unknown mutations.

bioinformatics↗

bioRxiv2025

Maternal immune activation imprints a regulatory T cell deficiency in offspring that drives an autism-like phenotype

Maternal immune activation (MIA) triggers an IL-17-driven autism spectrum disorder (ASD) in mouse and human offspring. While regulatory T cells (Tregs) regulate Th17 cells, their involvement in MIA and ASD pathogenesis is unknown. At the maternal level, we show that Treg stimulation suppresses interleukin-17 (IL-17) production and prevents ASD-like behaviors in offspring. At the offspring level, we show that MIA imprints a systemic and brain Treg deficiency, as evidenced by alterations in the methylome, transcriptome, and functional assays. This deficiency promotes brain inflammation, characterized by infiltration of IL-17-producing cells and neutrophils into the meninges and alterations in cortical brain structure. Stimulation of offspring Tregs with interleukin-2 reversed brain inflammation and cured established ASD-like behaviors. Thus, MIA-induced ASD is a neuroimmune disorder that can be reversed by immunomodulation. One sentence abstractMaternal immune activation during pregnancy imprints a Treg deficiency in offspring that perpetuates brain inflammation and an autism-like phenotype, which can be reversed by Treg stimulation.

immunology↗

bioRxiv2024

GDFold2: a fast and parallelizable protein folding environment with freely defined objective functions

An important step of mainstream protein structure prediction is to model the 3D protein structure based on the predicted 2D inter-residue geometric information. This folding step has been integrated into a unified neural network to allow end-to-end training in state-of-the-art methods like AlphaFold2, but is separately implemented using the Rosetta folding environment in some traditional methods like trRosetta. Despite the inferiority in prediction accuracy, the conventional approach allows for the sampling of various protein conformations compatible with the predicted geometric constraints, partially capturing the dynamic information. Here, we propose GDFold2, a novel protein folding environment, to address the limitations of Rosetta. On the one hand, GDFold2 is highly computationally efficient, capable of accomplishing multiple folding processes in parallel within the time scale of minutes for generic proteins. On the other hand, GDFold2 supports freely defined objective functions to fulfill diversified optimization requirements. Moreover, we propose a quality assessment (QA) model to provide reliable prediction on the quality of protein structures folded by GDFold2, thus substantially simplifying the selection of structural models. GDFold2 and the QA model could be combined to investigate the transition path between protein conformational states, and the online server is available at https://structpred.life.tsinghua.edu.cn/server_gdfold2.html.

bioinformatics↗

bioRxiv2021

HIV-specific CD8 T cells from elite controllers have an epigenetic imprint that preserves effector functions

Several lines of evidence support a central role for CD8 T cells as key determinants in the control of HIV, particularly in rare "elite controllers" who control the virus to undetectable levels in the blood in the absence of antiretroviral therapy (ART). While HIV-specific CD8 T cells isolated from elite controllers have enhanced antiviral cytokine production and proliferative capacity in response to antigen stimulation when compared to cells isolated from viremic or even aviremic ART-suppressed non-controllers, the cell-intrinsic mechanisms underlying the enhanced T cell memory-like function of HIV-specific CD8 T cells in elite controllers remain largely undefined. To identify the transcriptional and epigenetic pathways that regulate functional capacity in HIV-specific CD8 T cells in elite controllers, we performed genome-wide transcriptional and DNA methylation analysis of MHC Class I multimer+ CD8 T cells sorted from aviremic elite controllers compared to aviremic non-controllers on suppressive ART. Co-omics analysis revealed enrichment for gene signatures that support a multipotent differentiation state, cell survival, and a long-lived effector cell fate in HIV-specific CD8 T cells from elite controllers. Specifically, we observed DNA methylation programs at the transcription factor binding sites of the stem-associated factors TCF-1 and LEF1 that delineate HIV-specific CD8 T cells from elite controllers versus ART-treated individuals. HIV-specific CD8 T cells in elite controllers also maintain T cell receptor and IL-12/STAT4 pathway signaling and have suppressed pro-apoptotic TNF pathway signaling. These findings show that HIV-specific CD8 T cells from elite controllers have enhanced expression and DNA methylation programs that maintain developmental potential and in turn promote long-term survival, proliferative potential, and effector capacity. These data also provide new insights into the relationship between stem-associated transcription factors and stable epigenetic restriction of T cell developmental capacity.

immunology↗

bioRxiv2021

Maturation of Human Long-lived Plasma Cells Results in Resistance to Apoptosis by Transcriptional and Epigenetic Regulation

Antibody secreting cells (ASC) circulate after vaccination and migrate to the bone marrow (BM) where a subset known as long-lived plasma cells (LLPC) persist and secrete antibodies for a lifetime. The mechanisms of how circulating ASC become LLPC are not well elucidated. Here, we show that human blood ASCs have distinct morphology, transcriptomes, and epigenetics compared to BM LLPC. LLPC acquire transcriptional and epigenetic changes in the apoptosis pathway to support their survival. Upregulation of pro-survival gene expression accompanies downregulation of pro-apoptotic gene expression in LLPC. While pro-apoptotic gene loci are less accessible, pro-survival gene loci are not always accompanied by accessibility changes. Importantly, we show similar LLPC morphological and transcriptional maturation of blood ASC in response to the novel in vitro BM mimetic. In all, our study demonstrates that blood ASC in the BM microniche must undergo morphological and molecular changes to mature into apoptotic-resistant LLPC.

immunology↗

bioRxiv2019

Fcrl5 and T-bet define influenza-specific memory B cells that predict long-lived antibody responses

Seasonal influenza vaccination elicits hemagglutinin (HA)-specific CD27+ memory B cells (Bmem) that differ in expression of T-bet, BACH2 and TCF7. T-bethiBACH2loTCF7lo Bmem are transcriptionally similar to effector-like memory cells while T-betloBACH2+TCF7+ Bmem exhibit stem-like central memory properties. T-bethi Bmem do not express plasma cell-specific transcription factors but do exhibit transcriptional, epigenetic, metabolic and functional changes that poise the cells for antibody production. Consistent with these changes, D7 HA+ T-bethi Bmem express intracellular immunoglobulin and T-bethi Bmem differentiate more rapidly into ASCs in vitro. The T-bethi Bmem response positively correlates with long-lived humoral immunity and clonotypes from T-bethi Bmem are represented in the early secondary ASC response to repeat vaccination, suggesting that this effector-like population can be used to predict vaccine durability and recall potential.

immunology↗