Sijie Chen

Stanford, CA

I am a postdoctoral scholar in Prof. Lei Xing’s lab at Stanford University, working at the intersection of machine learning and computational biology. I build representations and generative models that preserve the structure of complex biological data, drawing on optimal transport, geometric deep learning, and cross-domain alignment.

My current work includes TransMap, which converts gene-expression profiles into image-native representations, and Dynode, which models 3D organ development with equivariant neural differential equations. I also contributed to scMulan, a large generative foundation model for single-cell analysis.

Before Stanford, I completed my PhD at Tsinghua University, advised by Prof. Xuegong Zhang and Prof. Michael S. Waterman. In hECA, I co-led the governance and harmonization of 1.09 million cells from 116 studies across 38 organs, designing a unified framework for heterogeneous metadata, hierarchical annotations, and cloud-based retrieval. My doctoral research also covered single-cell data integration and statistical methods for genomic sequence analysis.

I am open to industry research / engineering positions in machine learning, foundation models for science, and AI for medicine. CV (PDF) · GitHub · Google Scholar · LinkedIn

selected publications

NeurIPS

TransMap: Image-Native Representations for Single-Cell Genomics

Sijie Chen and others

NeurIPS (under review), 2026

Abs

Single-cell RNA-seq profiles are sparse, high-dimensional gene-expression vectors whose gene order carries no biological geometry. We introduce TransMap, a representation framework that converts each cell into a 2D transcriptomic image by arranging genes on a fixed grid via Gromov–Wasserstein optimal transport, so that nearby pixels correspond to genes with similar relational profiles. On five scRNA-seq benchmarks, CNN encoders on TransMap images match or outperform parameter-matched MLPs on scIB embedding metrics. A shared pixel grid further aligns cell states on a human–mouse pancreas benchmark without explicit ortholog conversion.
iScience

hECA: The Cell-Centric Assembly of a Cell Atlas

Sijie* Chen, Yanting* Luo, Haoxiang* Gao, and 8 more authors

iScience, 2022

HTML
Bioinformatics

A New Statistic for Efficient Detection of Repetitive Sequences

Sijie Chen, Yixin Chen, Fengzhu Sun, and 2 more authors

Bioinformatics, 2019

HTML
RECOMB/ISMB

scMulan: A Multitask Generative Pre-trained Language Model for Single-Cell Analysis

Haiyang Bian, Yixin Chen, Xiaomin Dong, and 7 more authors

In RECOMB / ISMB, 2024