Jianan Fan

Jianan Fan

Auto-generated profile, based on publicly available information

Jianan Fan

School of Computer Science, The University of Sydney, Sydney, NSW, Australia

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

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

Representative Publications

View All Publications
29 August 2025 in medRxiv
Verified
Preprint
Open access
PDF file
AXAngli Xue
Angli Xue
JFJianan Fan
Jianan Fan
ODOscar A. Dong
Oscar A. Dong
HHHao Lawrence Huang
Hao Lawrence Huang
LCLing Chen
Ling Chen
JPJoseph E. Powell
Joseph E. Powell
Understanding how genetic variation influences gene regulation at the single-cell level is crucial for elucidating the mechanisms underlying complex diseases. However, limited large-scale single-cell multi-omics data have constrained our understanding of the regulatory pathways linking variants to cell type-specific gene expression. Here we present chromatin accessibility profiles from 3.5 million peripheral blood mononuclear cells (PBMCs) across 1,042 donors, generated using single-cell ATAC-seq, including ∼100,000 PBMCs from 90 donors using multiome (RNA+ATAC) sequencing, with matched whole-genome sequencing. We identified 440,996 chromatin peaks across 28 immune cell types and mapped 243,225 chromatin accessibility quantitative trait loci (caQTLs), of which 60% were cell type-specific. Rare variant analysis revealed associations with 27,927 peaks, recapitulating the same cell type-specific regulatory architecture. Colocalization with the eQTLs from scRNA-seq data (5.4 million PBMCs) identified 31,688 candidate cis -regulatory elements; around half (44.40%) show evidence of putative mediating effects via chromatin accessibility. Integrating caQTLs with GWAS summary statistics for 16 diseases and 44 blood traits uncovered 4.5% - 22.6% more colocalized signals compared with using eQTLs alone, many of which have not been reported in prior studies. We show that the limited concordance between GWAS and eQTL signals can be substantially improved by incorporating cell type– and tissue-specific context, distal regulatory effects, models allowing for multiple causal variants, and promoter/enhancer priming, together increasing overlap from ∼20% to ∼60%. In addition, using a graph neural network, we inferred peak-to-gene relationships from unpaired multiome data by integrating caQTL and eQTL signals, achieving up to 80% higher accuracy than with paired multiome data lacking QTL information. This gain translated into improved gene regulatory network inference, enabling the identification of 128 additional transcription factors (TF)–target gene pairs (a 22% increase). Together, these results provide a comprehensive single-cell map of chromatin accessibility and genetic variation in human circulating immune cells, establishing a powerful resource for dissecting cell type-specific regulation and advancing our understanding of genetic risk for complex diseases.

29 August 2025 in medRxiv
Verified
Preprint
Open access
PDF file
AHAlbert Henry
Albert Henry
ASAnne Senabouth
Anne Senabouth
RTRika Tyebally
Rika Tyebally
BBBlake Bowen
Blake Bowen
PAPeter C. Allen
Peter C. Allen
JPJoseph E. Powell
Joseph E. Powell
Genome-wide association studies (GWAS) have been instrumental in uncovering the genetic basis of complex traits. When integrated with expression quantitative trait loci (eQTL) mapping, they can help identify effector genes within GWAS loci through gene regulatory mechanisms. Recent single-cell eQTL (sc-eQTL) studies suggest that genetic effects on gene expression are often cell type- and subtype-specific, offering opportunities to uncover context-specific roles of effector genes that traditional GWAS and eQTL studies in bulk tissues may obscure. Here, we leverage sc-eQTL mapping results from the TenK10K phase 1 project to prioritise effector genes across 69 diseases and 31 biomarker traits, from 12,266 eGenes identified in 28 peripheral immune cell types. Using Mendelian randomisation (MR) and colocalisation, we present a catalogue of predicted effector genes and their estimated directional effects on the trait at cell type resolution, spanning 54,245 associations across 5,270 genes and 28 cell types for disease traits, and 348,103 associations across 9,581 genes and 28 cell types for biomarker traits. By quantifying polygenic enrichment at both the single-cell and cell type levels, we identify distinct immune cell contributions to immune-related and systemic conditions. We demonstrate differential polygenic enrichment of Crohn’s disease and COVID-19 amongst dendritic cell subtypes. Integrating clinical drug development data shows that therapeutic compounds targeting the gene-trait associations identified in this study are twice as likely to secure regulatory approval. Using Crohn’s disease as a motivating example, we demonstrate how population-based sc-eQTL data can pinpoint risk loci, effector genes, and cell types, complementing findings from disease-focused tissue samples. Our findings provide a foundational resource for understanding the cell-type-specific genetic architecture of disease and for guiding therapeutic discovery.

13 November 2025 in IEEE Transactions on Medical Imaging
Verified
Journal Article
Closed access
TWTianyi Wang
Tianyi Wang
JFJianan Fan
Jianan Fan
DZDingxin Zhang
Dingxin Zhang
DLDongnan Liu
Dongnan Liu
YXYong Xia
Yong Xia
WCWeidong Cai
Weidong Cai
Histopathology and transcriptomics are fundamental modalities in cancer diagnostics, encapsulating the morphological and molecular characteristics of the disease. Multi-modal self-supervised learning has demonstrated remarkable potential in learning pathological representations by integrating diverse data sources. Conventional multi-modal integration methods primarily emphasize modality alignment, while paying insufficient attention to retaining the modality-specific intrinsic structures. However, unlike conventional scenarios where multi-modal inputs often share highly overlapping features, histopathology and transcriptomics exhibit pronounced heterogeneity, offering orthogonal yet complementary insights. Histopathology data provides morphological and spatial context, elucidating tissue architecture and cellular topology, whereas transcriptomics data delineates molecular signatures through quantifying gene expression patterns. This inherent disparity introduces a major challenge in aligning these modalities while maintaining modality-specific fidelity. To address these challenges, we present MIRROR, a novel multi-modal representation learning framework designed to foster both modality alignment and retention. MIRROR employs dedicated encoders to extract comprehensive feature representations for each modality, which is further complemented by a modality alignment module to achieve seamless integration between phenotype patterns and molecular profiles. Furthermore, a modality retention module safeguards unique attributes from each modality, while a style clustering module mitigates redundancy and enhances disease-relevant information by modeling and aligning consistent pathological signatures within a clustering space. Extensive evaluations on The Cancer Genome Atlas (TCGA) cohorts for cancer subtyping and survival analysis highlight MIRROR’s superior performance, demonstrating its effectiveness in constructing comprehensive oncological feature representations and benefiting the cancer diagnosis. Code is available at https://github.com/TianyiFranklinWang/MIRROR.

28 October 2024
Verified
Conference Paper
Open access
MSMingyu Sheng
Mingyu Sheng
JFJianan Fan
Jianan Fan
DLDongnan Liu
Dongnan Liu
RKRon Kikinis
Ron Kikinis
WCWeidong Cai
Weidong Cai
Surgical instrument segmentation (SIS) on endoscopic images stands as a long-standing and essential task in the context of computer-assisted interventions for boosting minimally invasive surgery. Given the recent surge of deep learning methodologies and their data-hungry nature, training a neural predictive model based on massive expert-curated annotations has been dominating and served as an off-the-shelf approach in the field, which could, however, impose prohibitive burden to clinicians for preparing fine-grained pixel-wise labels corresponding to the collected surgical video frames. In this work, we propose an unsupervised method by reframing the video frame segmentation as a graph partitioning problem and regarding image pixels as graph nodes, which is significantly different from the previous efforts. A self-supervised pre-trained model is firstly leveraged as a feature extractor to capture high-level semantic features. Then, Laplacian matrixs are computed from the features and are eigendecomposed for graph partitioning. On the "deep" eigenvectors, a surgical video frame is meaningfully segmented into different modules such as tools and tissues, providing distinguishable semantic information like locations, classes, and relations. The segmentation problem can then be naturally tackled by applying clustering or threshold on the eigenvectors. Extensive experiments are conducted on various datasets (e.g., EndoVis2017, EndoVis2018, UCL, etc.) for different clinical endpoints. Across all the challenging scenarios, our method demonstrates outstanding performance and robustness higher than unsupervised state-of-the-art (SOTA) methods. The code is released at https://github.com/MingyuShengSMY/GraphClusteringSIS.git.

8 May 2025
Verified
Conference Paper
Open access
JLJieting Long
Jieting Long
ZLZhiyuan Li
Zhiyuan Li
JFJianan Fan
Jianan Fan
ZLZhuonan Liang
Zhuonan Liang
AMAo Ma
Ao Ma
WCWeidong Cai
Weidong Cai

8 May 2025
Verified
Conference Paper
Open access
AMAo Ma
Ao Ma
ZLZhiyuan Li
Zhiyuan Li
ZLZhuonan Liang
Zhuonan Liang
TGTiancheng Gu
Tiancheng Gu
JFJianan Fan
Jianan Fan
WCWeidong Cai
Weidong Cai