Huayuan SmartGene Secures Tens of Millions RMB Seed Round to Advance AI-Powered Virtual Cell Platform for Human Drug Efficacy Prediction
核心洞察
Huayuan SmartGene (搜索) completed a seed round of tens of millions of RMB led by Tsinghua University Shuimu Venture Capital (搜索) to advance its AI virtual cell technology.
The company's Wise-Perturb platform uses single-cell multi-omics data to predict human drug responses, addressing the 90% failure rate in clinical trials due to animal-human physiological differences.
Validated in real-world studies, the model accurately predicted DS-8201 response in ovarian cancer (搜索) PDX models and stratified osimertinib efficacy in non-small cell lung cancer (搜索) patients.
Huayuan SmartGene (搜索) (华源智因), an AI virtual cell (AIVC) enterprise, has completed a seed round of financing totaling tens of millions of RMB, led by Tsinghua University Shuimu Venture Capital (搜索). The newly raised capital will be directed toward iterating the company's underlying multi-modal sequencing technologies, expanding collaborations with top-tier Class A tertiary hospitals, and growing its team. The company has already begun planning a subsequent financing round.
The founding team brings together senior pharmaceutical industry practitioners and computational biology researchers, supported by a scientific advisory board that includes experts from the Shenzhen National Gene Bank. CEO Du Runshi, a University of California, Los Angeles alumnus and serial AI entrepreneur, leads the company alongside CTO Wang Yixuan, a doctoral candidate in the Department of Computer Science and Engineering at the Chinese University of Hong Kong who previously led development of the xTrimoSC Perturb model at BioMap. Li Yu, an assistant professor in the same department, serves as Chief AI Scientist.
Addressing the Clinical Translation Gap
The pharmaceutical industry's well-known "10 years, 1 billion dollars" dilemma reflects a stark reality: 90% of R&D resources are consumed during human clinical trials, yet only about 10% of candidate drugs ultimately gain market approval. A core driver of this failure rate is the profound disconnect between animal model physiology and actual human biology.
"Most traditional AI pharmaceutical companies focus on solving the front-end problems of R&D — using AI to more efficiently identify potential disease-causing targets and generate compounds," said Du Runshi. "However, this is only optimizing the first step of 'drug discovery,' which cannot fully predict whether the drug will be effective or have side effects when applied to the human body."
The Wise-Perturb Platform: Multi-Omics Cellular Modeling
Huayuan SmartGene (搜索)'s core technology, the Wise-Perturb cellular drug perturbation application model, focuses on predicting human drug efficacy by simulating how drugs affect the coordinated activity of approximately 20,000 protein-coding genes within human cells. The platform addresses two critical shortcomings of traditional AI cell models.
First, conventional models rely heavily on single RNA sequencing data derived from artificially modified immortal cell lines — cells divorced from their original pathological microenvironment and lacking patient-specific genetic characteristics. Moreover, traditional sequencing cannot simultaneously capture DNA, RNA, and protein information, making it difficult to reconstruct authentic cellular regulatory logic.
To overcome these limitations, Huayuan SmartGene (搜索) has built a single-cell multi-omics integrated analysis system that captures cellular DNA, transcriptome, and proteome data from the same cell. A self-developed multi-modal fusion algorithm connects these three omics layers into a pyramid-structured database: a bottom layer of massive static single-cell sequencing data, a middle layer containing hundreds of millions of paired drug and gene perturbation experiments from in vitro cell lines and PDX models, and a top layer of scarce paired sequencing data from human clinical tumor cohorts before and after drug administration.
"We have reached an innovative cooperation with top-tier Class A tertiary hospitals to build a joint laboratory," Du Runshi noted. "We hope to achieve in-depth cooperation with at least 30 top Class A tertiary hospitals in the next 1-3 years."
Second, Wise-Perturb incorporates a cell type recognition architecture that accounts for physiological differences across cell types — acknowledging, for instance, that a drug effective in the lungs may damage the liver. This design enables zero-shot generalization prediction capabilities across different cells, cancer types, and new drugs, allowing the model to predict human effects for novel diseases or drugs using only limited known patient data, without requiring large-scale retraining for each indication.
Clinical Validation
Huayuan SmartGene (搜索) provided two real-world verification cases. In the first, the model was trained on clinical data from breast cancer (搜索) patients and then used to predict drug response in ovarian cancer (搜索) PDX models for the antibody-drug conjugate DS-8201 — without any ovarian cancer training samples. The model scores matched real drug efficacy observed in animals, with improved prediction consistency compared to general virtual cell models.
In a second study conducted with the Cancer Hospital of the Chinese Academy of Medical Sciences, the model performed drug efficacy stratification for the targeted non-small cell lung cancer (搜索) drug osimertinib. "The hospital provided the baseline tumor sequencing data of patients before treatment, without releasing any clinical follow-up outcomes. The model independently output the patient's drug efficacy stratification results, which were consistent with the long-term clinically observed treatment effects," Du Runshi said.
Commercial Traction and Strategy
This clinically validated prediction capability has enabled Huayuan SmartGene (搜索) to generate revenue from its earliest stages. The company has "received cooperation orders from many top Class A tertiary hospitals, multinational pharmaceutical companies, and AI biotechnology enterprises, and the service fees for multiple projects have been fully collected," according to Du Runshi.
The company currently offers two core commercial services: customized preclinical pipeline value assessment and stratified patient selection for clinical trials for pharmaceutical companies, and joint wet-lab and dry-lab collaborations with Class A tertiary hospitals for projects including drug repurposing and novel target screening. Longer-term plans include a joint R&D model that shares risks with pharmaceutical partners and distributes revenue after drugs reach the market.
"We prioritize the implementation of small-scale verification projects, use reproducible and highly consistent prediction data to build customer trust, and then continue to deepen long-term cooperation," Du Runshi concluded. "At present, human precision drug efficacy prediction is a rigid demand in the industry, and there is a lack of mature alternative technical solutions."