Umair Khan

Umair Khan

Auto-generated profile, based on publicly available information

Umair Khan

Bakar Computational Health Sciences Institute, UCSF, San Francisco, USA

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

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

Representative Publications

View All Publications
13 May 2021 in The Journal of Physiology
Verified
Journal Article
Closed access
PDF file
BYBenny Yue
Benny Yue
BHBassam G. Haddad
Bassam G. Haddad
UKUmair Khan
Umair Khan
HCHonghong Chen
Honghong Chen
MAMena Atalla
Mena Atalla
DBDonglin Bai
Donglin Bai

23 October 2025 in bioRxiv
Verified
Preprint
Open access
SGShreshth Gandhi
Shreshth Gandhi
FJFarnoosh Javadi
Farnoosh Javadi
VSValentine Svensson
Valentine Svensson
UKUmair Khan
Umair Khan
MJMatthew G. Jones
Matthew G. Jones
NANima Alidoust
Nima Alidoust
Foundation models have transformed natural language processing and computer vision, yet their potential in single-cell biology—particularly for complex diseases such as cancer— remains underexplored. We present Tahoe-x1 (Tx1), a family of perturbation-trained single-cell foundation models with up to 3 billion parameters. Tx1 is pretrained on large-scale single-cell transcriptomic datasets, including the Tahoe-100M perturbation compendium, and fine-tuned for cancer-relevant tasks. Through architectural optimizations, data loader refinements, and efficient training strategies, Tx1 achieves 3–30× higher compute efficiency than prior implementations of cell-state models. Tx1 jointly learns representations of genes, cells, and compounds using a masked-expression generative objective that incorporates a drug token, enabling flexible adaptation to diverse downstream applications. We evaluate Tx1 across four key disease-relevant benchmarks: (1) prediction of overall and context-specific gene essentiality, (2) identification of genes contributing to the hallmarks of cancer, (3) cell-type classification, and (4) prediction of perturbation responses in held-out cellular contexts. Tx1 achieves state-of-the-art performance across all tasks. We release pretrained checkpoints, training code, and evaluation workflows to accelerate the development of perturbation-trained single-cell foundation models for applications in precision oncology and beyond.

1 December 2021 in Biophysical Journal
Verified
Journal Article
Closed access
JTJun-Jie Tong
Jun-Jie Tong
UKUmair Khan
Umair Khan
BHBassam G. Haddad
Bassam G. Haddad
PMPeter J. Minogue
Peter J. Minogue
EBEric C. Beyer
Eric C. Beyer
LELisa Ebihara
Lisa Ebihara

31 July 2025 in Cell Reports Medicine
Verified
Journal Article
Open access
UKUmair Khan
Umair Khan
TOTomiko T. Oskotsky
Tomiko T. Oskotsky
BYBahar D. Yilmaz
Bahar D. Yilmaz
JRJacquelyn Roger
Jacquelyn Roger
KGKetrin Gjoni
Ketrin Gjoni
MSMarina Sirota
Marina Sirota
Endometriosis is a prevalent, complex, inflammatory condition associated with a diverse range of symptoms and comorbidities. Despite its substantial burden on patients, population-level studies that explore its comorbid patterns and heterogeneity are limited. In this retrospective case-control study, we analyze comorbidities from over forty thousand endometriosis patients across six University of California medical centers using de-identified electronic health record (EHR) data. We find hundreds of conditions significantly associated with endometriosis, including genitourinary disorders, neoplasms, and autoimmune diseases, with strong replication across datasets. Clustering analyses identify patient subpopulations with distinct comorbidity patterns, including psychiatric and autoimmune conditions. This study provides a comprehensive analysis of endometriosis comorbidities and highlights the heterogeneity within the patient population. Our findings demonstrate the utility of EHR data in uncovering clinically meaningful patterns and suggest pathways for personalized disease management and future research on biological mechanisms underlying endometriosis.

12 June 2026 in bioRxiv
Verified
Preprint
Open access
VSValentine Svensson
Valentine Svensson
UKUmair Khan
Umair Khan
HHHamed Heydari
Hamed Heydari
AUAirol A. Ubas
Airol A. Ubas
NTNicole Thomas
Nicole Thomas
SGShreshth Gandhi
Shreshth Gandhi
Predicting how cells, tissues, and patients will respond to a drug, cytokine, or genetic perturbation is central to biological and clinical reasoning. The practical goal is to estimate the analysis-ready readouts that support this reasoning: which cellular responses are context-dependent, which perturbations reveal shared or divergent mechanisms, and which observations should become the basis for the next experiment. Here we introduce Rhaister, a perturbation-response predictor that operates directly on screen-level summary statistics. By measuring just a few perturbations in a new biological context, Rhaister predicts the unmeasured perturbations by learning how response patterns vary across reference contexts. This formulation applies to both fine-grained molecular readouts, such as transcriptional responses from Tahoe-100M or other large perturbation screen, or on phenotypic endpoints. To train and apply Rhaister on phenotypic endpoints we created Emerald Bay, a purpose-built Tahoe dataset that unifies multi-day cancer drug perturbation, pooled Mosaic tumor contexts, and paired transcriptomic response measurements*. Across these settings, Rhaister matches or exceeds substantially more expensive virtual-cell models, often achieving the highest values possible in evaluation metrics, while training in seconds and running predictions in milliseconds. On Emerald Bay, Rhaister predicts context-specific drug phenotypes from sensitivity measurements alone and improves further when including transcriptomic information. We also introduce Rhaister-O, predicting drug responses in new contexts from baseline expression alone and, to our knowledge, provides the first zero-shot model for this task. Rhaister establishes summary-statistic perturbation modeling as a fast, interpretable framework for predicting biological response across new contexts. *: https://huggingface.co/collections/tahoebio/rhaister

25 April 2025 in medRxiv
Verified
Preprint
Open access
PDF file
YKYukari Katsuhara
Yukari Katsuhara
UKUmair Khan
Umair Khan
ZMZachary A. Miller
Zachary A. Miller
IAIsabel E. Allen
Isabel E. Allen
TOTomiko T. Oskotsky
Tomiko T. Oskotsky
ATAlice S. Tang
Alice S. Tang
Summary: We applied unsupervised learning techniques to electronic medical records from UCSF to identify distinct Alzheimer’s disease subgroups based on comorbidity profiles. Given the well-known female sex predominance in Alzheimer’s disease prevalence, we performed sex-stratified analyses to evaluate differences in disease manifestations based on sex. Findings were validated using an independent UC-Wide dataset. Among 8,363 patients, we identified five Alzheimer’s disease subphenotypes, characterized by comorbidities related to cardiovascular conditions, gastrointestinal disorders, and frailty-related conditions such as pneumonia and pressure ulcers. We further refined significant comorbidity variations across clusters through sex-stratified analyses, observing a higher prevalence of circulatory diseases in males in Cluster 2 and bladder stones in females in Cluster 3. Key results were consistent across the UCSF and UC-Wide datasets. Our study identifies clinically meaningful Alzheimer’s disease subgroups, along with sex-specific variations, suggesting underlying biological factors, and indicates the potential utility of these findings in informing individualized therapeutic regimens.