Yin Fang

I recently earned my Ph.D. degree in Computer Science from Zhejiang University in June 2024, under the supervision of Professors Huajun Chen, Ningyu Zhang and Xiaohui Fan. I have joined the National Institutes of Health (NIH) as a postdoctoral fellow, with Dr. Zhiyong Lu as my mentor.

My research primarily focuses on AI4Bioinformatics. I am interested in using large language models and graph neural networks to analyze biomedical data, e.g. molecules, proteins, and single-cell data. Feel free to reach me if you are interested in my research!

Email  /  Google Scholar  /  DBLP  /  Github  /  Twitter

KANO is a Knowledge graph-enhanced molecular contrAstive learning framework with fuNctional prOmpt, designed to exploit fundamental domain knowledge in both pre-training and fine-tuning phases.

Mol-Instructions is a Large-Scale Biomolecules Instruction Dataset for Large Language Models.

MolGen is a pre-trained molecular language model tailored specifically for molecule generation.

KCL is a novel Knowledge-enhanced Contrastive Learning framework for molecular representation learning, built upon a Chemical Element Knowledge Graph (KG) that summarizes microscopic associations between elements.

Bulk2Space is a two-step spatial deconvolution method based on deep learning frameworks, which converts bulk transcriptomes into spatially resolved single-cell expression profiles.

scSpace (single-cell and spatial position associated co-embeddings) is an integrative algorithm that integrates spatial transcriptome data to reconstruct spatial associations of single cells within scRNA-seq data.

This paper proposes possible directions for future molecular science research. These suggestions aim to forge new paths for exploring the intricacies of molecular structures, potentially bringing new efficiencies and innovations in the field.

DeepSorption is a spatial atom interaction learning network that realizes accurate, fast, and direct structure-adsorption prediction with only information of atomic coordinate and chemical element types.

This paper offers a comprehensive review of molecular learning, with a focus on the knowledge-informed paradigm transfer approach. It provides an overview of the various paradigms and their technical solutions, as well as a summary of the external domain knowledge used to guide the transfer process for each molecular learning task.