Poster Presentations 2026

Xiaofeng Tan, Ph.D.,DeepCovalent, Inc.

Discovering New Drugs in a Single Step via GenAI

OS3D™, a GenAI framework for de novo drug discovery, reduces early-stage discovery from 2–4 years to ~3 weeks via direct multi-objective molecular generation. The platform achieved 81% re-discovery of 490 FDA-approved drugs and generated GLP-1R candidates with favorable pharmacological profiles versus industry reference compounds, supporting accelerated and higher-success therapeutic discovery.

Yi Ni, Liwei Zhu, Shuai Li, Bio LIMS INC 

Multi-agent AI System for Autonomous CAR-T Development: Integrated Target Discovery, Toxicity Prediction, and Rational Molecular Design for Cancer Immunotherapy   

It introduces a multi-agent AI System (Target Selection, Toxicity Prediction, Molecular Design, Patent Intelligence, Clinical Translation, Decision Synthesis) orchestrate public databases (GTEx, HPA, FAERS, PubMed, OpenTargets ..), specialized tools (ESM, AlphaFold, DiffDock, RD..), and other systems.

Zhenyu Wu, Qingbo Xu, Ellen Wan, Hongbo Zhang, HitChem Ltd

From Design to Application: A CRBN Molecular Glue Library Featuring Diverse and Unique CRBN Binders

In this study, we designed and identified several novel CRBN binders exhibiting excellent or acceptable binary binding affinities. Compared with previously reported CRBN ligands, our compounds possess unique structural features and enhanced hydrophobic properties. Derivatives based on these scaffolds enabled the CRBN molecular glue library extend into previously unexplored chemical space,

Angie Yu, Mark Chen, Altruist Biologics

Addressing Challenges of Complex Molecules With a CMC Strategy Framework

The development of bsAbs, msAbs, and Fc-fusion proteins presents unique CMC challenges due to their structural complexity and heterogeneity. Altruist Biologics addresses these challenges through advanced cell line development, process optimization, analytical characterization, and purification technologies.

Menglin Kong, School of Pharmacy, Massachusetts College of Pharmacy and Health Sciences

Beyond the Numeric Rating Scale: A Systematic Review of Pain Assessment in Clinical Trials and a Proposal for Multidimensional Patient-Reported Measurement

This systematic review evaluated pain assessment methods in clinical trials. Results showed dominant use of unidimensional tools like VAS and NRS, which may inadequately capture pain complexity. A novel multidimensional patient-reported framework is proposed to improve assessment precision and patient engagement.

Shangyuan Cui, Zhongfeng Ye, Mariah L. Arral, Lihan Liu, Benson Weng, Xiaohan Zhang, Qiaobing Xu, and David L. Kaplan, Department of Biomedical Engineering, Tufts University

Silk-Based Formulation Component Enhances mRNA-LNP Vaccine Efficacy and Storage Stability at Non-frozen Temperatures

Lipid nanoparticles (LNPs) are clinically validated carriers for nucleic acid therapeutics; however, achieving targeted delivery to reticuloendothelial organs beyond the liver, lung, and spleen remains a major challenge. Here, we introduce a strategy for post-fabrication engineering of a custom protein corona on mRNA-LNPs using cationic silk fibroin (SF), a biocompatible and chemically tunable protein polymer. SF-coated LNPs exhibit enhanced cellular uptake and endosomal escape, resulting in a 3.6-fold increase in lymph node delivery and a 2.5-fold extension in in vivo protein expression compared to unmodified LNPs. In a cancer vaccine model, SF-LNPs significantly improve dendritic cell maturation, antigen cross-presentation, and cytotoxic T cell activation, leading to robust protection against tumor growth and metastasis, as well as durable immunological memory. Furthermore, we demonstrate that SF-LNPs have improved storage stability at non-frozen temperature by enhancing LNP colloidal stability. This work expands the formulation space for LNPs and establishes silk fibroin as a modular surface engineering tool for enhancing the efficacy, specificity, and stability of mRNA-based therapeutics.

Zeliang Guan, W. Frank An, Biocytogen Pharmaceuticals (Beijing) Co

Novel Blood-Brain Barrier Penetrating Heavy Chain-Only Antibodies Targeting CD98

The blood-brain barrier (BBB) limits drug delivery to the brain. Targeting receptor-mediated transcytosis offers a solution. We developed heavy chain-only antibodies (HcAbs) against CD98, highly expressed on brain endothelial cells. Four novel anti-CD98 HcAbs showed superior brain penetration in vivo, highlighting their promising potential as CNS therapeutic shuttles.

Jintian (Tim) Lyu, Gary Ma, Sam Xu, Taimei Technology

Building a Future-Ready Clinical R&D Model: How A Leading Multinational Pharmaceutical Company Is Partnering with Taimei to Transform Clinical Research in China

This case study evaluates an AI-enabled clinical R&D transformation model in China through a partnership between a leading multinational pharmaceutical company and Taimei Technology, integrating digital infrastructure, clinical operations, data management, pharmacovigilance, and AI workflows to improve cost efficiency and trial execution.

Eric Yao, Kodexia Platform Team, XtalPi

Kodexia™: An Integrated siRNA Design Platform Powered by Generative AI

Kodexia™ is an integrated siRNA de novo design platform powered by generative AI and dry–wet iterative optimization for sequence and chemical modification co-design. The platform combines physically grounded nucleic acid modeling, generative and discriminative AI models, proprietary high-quality datasets, and high-throughput wet-lab validation to accelerate the discovery and optimization of siRNA

Lee Dolat, Tyler Vincent, Diane Ignar, Yaguang Si, Josh Sommer, Daniel M. Freed, Chordoma Foundation

A preclinical assay pipeline to identify and optimize TBXT-targeted therapies in chordoma and other solid tumors

Chordoma is a rare bone cancer with no approved systemic therapy. Functional genomics studies identify TBXT as the most selectively essential gene in chordoma cells. We have established a preclinical assay pipeline to support the identification and optimization of TBXT-targeted therapies. This platform provides a foundation for advancing TBXT-directed drug discovery.

Haoran Wu, Lijun Wang, Yingying Xu, Yun He, Hongjiang Miao, Nona Biosciences

Fully Human HCAb Antibody Discovery Enhanced by Hu-mAtrix(TM) AI-Integrated Platform

Hu-mAtrlx"" is Nona Biosciences' proprietary AI-platform designed to support antibody discovery and early optimization within Harbor Mice®-based programs (fully human H2L2 and HCAb). Rather than replacing established and clinically proven discovery workflows, Hu-mAtrlx""functions as an integrated analytical layer, applying data-driven insights at key decision points to support target validation, candidate selection, prioritization, and risk mitigation. HCAb Harbor Mice® generate fully human antibodies with single-domain binders (VH domains)that exhibit high developability for next-generation therapies. To enhance sequence diversity in antibody discovery campaigns, Hu-mAtrlx"" combines proprietary Harbour Mice® HCAb sequence datasets, target-specific data, and several finely-tuned AI models. This integration facilitates the generation and classification of sequences, in silica developability assessment, and the selectionofstructure-aware binders within a seamless dry-to-wet workflow.

Shuang Cao', Rui Lil, Louise Liu1, Tom Chen2, Zhiyuan Cheng1, Kaige Zheng', Coco Qul, Vivian Zhang', Alexandre Duprey1 1 Hill Research, 2 Carnegie Mellon University

From Contraindications to Citations: Verifiable Medication-Safety QA

THE PROBLEM: A fluent answer can be dangerous if it misses what matters. HOW WE SOLVE IT: AegisRAG retrieves for what would be most dangerous to miss. AegisRAG treats retrieval as budgeted evidence selection for safety-facet coverage. REAL IMPACT: In a prospective A/B deployment with real clinicians, AegisRAG cut major medication-safety errors and shortened time-to-decision - while every answer stayed auditable back to its evidence.

Mengran Li, Haipeng Liu, Chengzhang Shang, Yi Yang, Biocytogen

BCG048, a Novel Bispecific Dual-payload ADC Targeting ITGB6 and B7H3, Exhibited Potent Efficacy in Patient-derived Tumor Xenograft Models

ITGB6 and B7H3 are co-overexpressed in multiple solid tumors and linked to poor prognosis, making them promising ADC targets. Using the RenLite platform, we developed a bispecific antibody targeting both proteins. The bsAb showed strong tumor binding, enhanced internalization, and potent in vivo efficacy. Dual-payload BCG048 outperformed benchmark ADCs in PDX models.

Ruili Lv, Suman Zhao, Qingqing Xu, Xiaofei Zhou, Biocytogen

FcRn Humanized Mice with an Immunodeficient Background Can be Used to Evaluate the Pharmacokinetics of Antibody Drugs while Avoiding Anti-Drug Antibodies

FcRn recycles IgG in a pH-dependent manner to prevent lysosomal degradation and extend antibody half-life, while Rag2 deficiency eliminates T and B cells, preventing ADA formation. We generated B-hFcRn, Rag2 KO mice expressing human FcRn but lacking lymphocytes. The mice showed human FcRn expression, absence of T/B cells, and enabled ADA-free PK evaluation of YTE antibodies with extended half-life

Philip Gorman, Gavin Zhang, Huaxing Zhu

Rapid, High-throughput antibody expression and characterization to accelerate AI/ML driven drug discovery and iterative model training

AI/ML-driven antibody discovery requires large, high-quality experimental datasets for model training. We present a high-throughput platform integrating antibody expression, SPR binding analysis, and developability profiling to generate standardized, AI-ready data that accelerates iterative lab-in-a-loop optimization.