Siyu Tao

Siyu Tao

Auto-generated profile, based on publicly available information

Siyu Tao

School of Information Science and Technology, ShanghaiTech University, Shanghai, China and Lingang Laboratory, Shanghai, China

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
6 October 2025 in IEEE Transactions on Computational Biology and Bioinformatics
Verified
Journal Article
Closed access
STSiyu Tao
Siyu Tao
YYYang Yang
Yang Yang
XLXin Liu
Xin Liu
YFYimiao Feng
Yimiao Feng
ZJZheng Jie
Zheng Jie
Predicting biomedical interactions is crucial for understanding various biological processes and drug discovery. Graph neural networks (GNNs) are promising in identifying novel interactions when extensive labeled data are available. However, labeling biomedical interactions is often time-consuming and labor-intensive, resulting in low-data scenarios. Furthermore, distribution shifts between training and test data in real-world applications pose a challenge to the generalizability of GNN models. Recent studies suggest that pre-training GNN models with self-supervised learning on unlabeled data can enhance their performance in predicting biomedical interactions. Here, we propose LukePi, a novel self-supervised pre-training framework that pre-trains GNN models on biomedical knowledge graphs (BKGs). LukePi is trained with two self-supervised tasks: topology-based node degree classification and semantics-based edge recovery. The former is to predict the degree of a node from its topological context and the latter is to infer both type and existence of a candidate edge by learning semantic information in the BKG. By integrating the two complementary tasks, LukePi effectively captures the rich information from the BKG, thereby enhancing the quality of node representations. We evaluate the performance of LukePi on two critical link prediction tasks: predicting synthetic lethality and drug-target interactions, using four benchmark datasets. In both distribution-shift and low-data scenarios, LukePi significantly outperforms 22 baseline models, demonstrating the power of the graph pre-training strategy when labeled data are sparse.


30 June 2026
Verified
Conference Paper
Open access
YLYuanxian Li
Yuanxian Li
STSiyu Tao
Siyu Tao
YFYimiao Feng
Yimiao Feng
ZJZheng Jie
Zheng Jie
Synthetic lethality (SL) has emerged as a clinically validated approach to identifying anticancer therapeutic targets, effectively exposing the critical vulnerabilities unique to cancer cells. While Large Language Models (LLMs) have achieved significant progress in scientific research, their reliability in reasoning about SL gene pairs remains unexplored. To fill this gap, we present a unified framework, named BASIS, for evaluating and improving LLMs in inferring SL interactions for anticancer drug target discovery. We first construct SL-Bench, a systematic benchmark encompassing seven hierarchical tasks, to rigorously evaluate LLMs across a spectrum of cognitive demands ranging from factual recall to clinical hypothesis formulation, thereby probing their capacity for deep mechanistic reasoning over shallow statistical correlations. Our benchmarking study revealed a critical limitation: contemporary LLMs suffer from substantial logical inconsistencies when inferring intricate biological dependencies. To resolve this issue, we propose SL-Agent, a neuro-symbolic Solver-Verifier architecture that integrates GraphRAG with iterative self-reflection, demonstrating consistent performance improvements when applied to various base models. Notably, on the most challenging task, an 8B open-source model equipped with SL-Agent outperforms leading proprietary models, exceeding GPT-5.2 in accuracy by 14.0 pp and Gemini-3-Pro in safety by 28.0 pp. By unifying systematic evaluation with targeted improvement, BASIS establishes a practical path toward reliable, AI-driven synthetic lethality discovery. The code is available in https://github.com/davidwushi1145/BASIS.