Xingzhong Zhao
Associate Research Scientist, Yale UniversityBiomedical Informatics & Data Science
I study how genetic variation shapes the human body as seen in medical images, with a focus on brain ageing and neurological disease. I use AI to turn brain and heart scans into heritable traits, then map those traits across the genome in large biobanks.
I build tools to make this possible at scale, including TorchGWAS, which runs genome-wide association tests for thousands of image-derived phenotypes on a GPU. My PhD in Biomedical Engineering at Fudan University covered neuroimaging genomics and gene regulation in brain disorders. Before that, I trained in computer science.
- 2026 – now
- Associate Research ScientistBiomedical Informatics & Data Science, Yale University
- 2026
- Faculty AssociateUTHealth Houston
- 2024 – 2025
- Postdoctoral FellowUTHealth Houston
- 2021 – 2024
- Ph.D., Biomedical EngineeringFudan University
- 2018 – 2021
- M.S., Biomedical EngineeringFudan University
From voxels to variants
Each project follows the same path: from images, to learned traits, to genetic associations, to biological mechanisms.
-
Images
Diffusion and structural brain MRI and cardiac MRI from large population biobanks.
-
Representations
Unsupervised and contrastive deep learning that turns each image into compact, heritable endophenotypes.
-
Associations
GPU-accelerated GWAS that tests thousands of phenotypes, including in samples with related individuals.
-
Mechanisms
Enhancer–promoter maps, cell type-specific Mendelian randomization and gene prioritization to explain the hits.
Selected work
Papers I led as first or co-first author. # marks equal contribution.
-
2026
Nature Communications
Genetic architecture of white matter microstructure captured by unsupervised deep representation learning of fractional anisotropy maps
Learns white matter traits directly from diffusion MRI with no hand-picked tracts, then maps their genetics.
-
2026
Nature Communications (accepted)
Uncovering genetic architecture of the heart via genetic association studies of unsupervised deep learning derived endophenotypes
Applies the same unsupervised approach to cardiac MRI and maps the genetics of the learned heart traits. Co-first author.
-
2026
Bioinformatics Advances · arXiv
TorchGWAS: GPU-accelerated GWAS for thousands of quantitative phenotypes
Runs association tests on the GPU so that thousands of image-derived phenotypes can be scanned in one pass.
-
2026
PLOS Computational Biology
Deciphering cell type-specific causal genetic effects on brain imaging-derived phenotypes and disorders with single-cell Mendelian randomization
Uses single-cell Mendelian randomization to find which cell types carry causal genetic effects on brain imaging traits and disorders. Co-first author.
-
2026
IEEE Transactions on Medical Imaging
EIGNN: An explainable imaging-genetic neural network for robust Alzheimer's disease risk prediction
Combines brain imaging and genetics in an interpretable network to predict Alzheimer's disease risk. Co-first author.
-
2025
Genomics, Proteomics & Bioinformatics
Regulatory genomic circuitry of brain age by integrative functional genomic analyses
Combines functional genomics data to trace the regulatory programs behind brain ageing.
-
2023
Genome Medicine
Prioritizing genes associated with brain disorders by leveraging enhancer–promoter interactions in diverse neural cells and tissues
Builds enhancer–promoter maps for 439 neural cell and tissue types and uses them to rank genes for brain disorders. Co-first author.
-
2021
Methods
Deep learning of brain magnetic resonance images: A brief review
A review of deep learning methods for brain MRI, from segmentation to disease prediction.
All publications
# co-first author, * corresponding author. Titles link to the paper DOI where available, otherwise to a Google Scholar search. See the full list on Google Scholar.
2026
- A review of self-supervised imaging phenotypes for genome-wide association studies
Communications Biology
- TorchGWAS2: Cost-effective phenome- and genome-wide association testing in related samples
medRxivPreprint
- PerturbLDM: conditional latent diffusion for modelling single-cell perturbation responses
bioRxivPreprint
- Deciphering cell type-specific causal genetic effects on brain imaging-derived phenotypes and disorders with single-cell Mendelian randomization
PLOS Computational BiologyCo-first author
- EIGNN: An explainable imaging-genetic neural network for robust Alzheimer's disease risk prediction
IEEE Transactions on Medical ImagingCo-first author
- Genetic architecture of white matter microstructure captured by unsupervised deep representation learning of fractional anisotropy maps
Nature CommunicationsFirst author
- Replicability of unsupervised deep learning derived image phenotypes
bioRxivPreprint
- Learning dynamics of unsupervised deep learning reveal epoch-specific genetic architectures of brain morphology
bioRxivPreprint
- Learning heritable multimodal brain representation via contrastive learning
Research SquarePreprint
- Uncovering genetic architecture of the heart via genetic association studies of unsupervised deep learning derived endophenotypes
Nature Communications (accepted)Co-first author
- TorchGWAS: GPU-accelerated GWAS for thousands of quantitative phenotypes
arXiv:2604.21095First authorPreprint
- A trispecific antibody engaging T cells with tumour and myeloid cells augments antitumour immunity
Nature Biomedical Engineering
- Impacts of host genetics on gut microbiome composition in Alzheimer's disease
Microbiome
- TorchGWAS 1.0: GPU-accelerated GWAS at scale
Bioinformatics AdvancesFirst author
- Sum-h2, enabling genetic discovery for deep learning-derived phenotypes through a fast evaluation framework and arena of performance
bioRxivPreprint
2025
- Unveiling the genetic architecture of white matter tracks through unsupervised deep representation learning of fractional anisotropy images
Alzheimer's & Dementia (AAIC abstract)First authorConference
- Regulatory genomic circuitry of brain age by integrative functional genomic analyses
Genomics, Proteomics & BioinformaticsFirst author
- Multimodal profiling identifies CD4+ CXCR5+ PD-1– Tfh cells as prognostic and predictive biomarkers in diffuse large B-cell lymphoma
Chinese Medical Journal
- Spatiotemporally resolved transcriptomics reveals the cellular dynamics of human retinal development
Nature Communications
- Leveraging deep learning to enhance MRI for brain disorders
medRxivPreprint
- Association between twin status with cognitive, behavioral development and brain structure in early adolescence: a retrospective cohort analysis based on the Adolescent Brain Cognitive Development study
European Child & Adolescent PsychiatryCo-first author
2024
- Association of birth weight with neuropsychological functioning in early adolescence: A retrospective cohort study
Psychiatry ResearchCo-first author
- Effects of open-skill and closed-skill exercise on subthreshold depression in female adolescents: a randomized controlled trial
International Journal of Clinical and Health Psychology
- STAB2: an updated spatio-temporal cell atlas of the human and mouse brain
Nucleic Acids Research
2023
- Prioritizing genes associated with brain disorders by leveraging enhancer-promoter interactions in diverse neural cells and tissues
Genome MedicineFirst author
- Improving Alzheimer's disease diagnosis with multi-modal PET embedding features by a 3D multi-task MLP-mixer neural network
IEEE Journal of Biomedical and Health Informatics
- Deciphering the genetic architecture of human brain structure and function: a brief survey on recent advances of neuroimaging genomics
Briefings in BioinformaticsFirst author
2022
- Low-dose aspirin in the prevention of preeclampsia in twin pregnancies: a real-world study
Frontiers in Cardiovascular MedicineCo-first author
2021
- Deep learning of brain magnetic resonance images: A brief review
MethodsFirst author
- Integration of multimodal data for deciphering brain disorders
Annual Review of Biomedical Data Science
- Identifying age-specific gene signatures of the human cerebral cortex with joint analysis of transcriptomes and functional connectomes
Briefings in BioinformaticsFirst author
- A 12-genus bacterial signature identifies a group of severe autistic children with differential sensory behavior and brain structures
Clinical and Translational Medicine
2020
2019
- Predicting drug-disease associations with heterogeneous network embedding
Chaos: An Interdisciplinary Journal of Nonlinear Science
- Deep learning based multi-modal cardiac MR image segmentation
STACOM Workshop (MICCAI)Corresponding authorConference
No publications match. Clear the search or switch the filter to All.
Software & data
TorchGWAS
GPU-accelerated GWAS for thousands of quantitative phenotypes. TorchGWAS2 adds support for related samples.
Read the paperBrain EPI atlas
Enhancer–promoter interactions for 439 neural cell and tissue types, plus prioritized genes for each brain disorder and cognitive trait.
Open the data portal