Devansh Pandey

Devansh Pandey

Auto-generated profile, based on publicly available information

Devansh Pandey

Department of Integrative Biology, The University of Texas at Austin, Austin, TX, USA

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
11 April 2025 in Science
Verified
Journal Article
Closed access
LXLiaoyi Xu
Liaoyi Xu
EKEucharist Kun
Eucharist Kun
DPDevansh Pandey
Devansh Pandey
JWJoyce Y. Wang
Joyce Y. Wang
MBMarianne Brasil
Marianne Brasil
VNVagheesh M Narasimhan
Vagheesh M Narasimhan
Human pelvic evolution following the human-chimpanzee divergence is thought to result in an obstetrical dilemma, a mismatch between large infant brains and narrowed female birth canals, but empirical evidence has been equivocal. By using deep learning on 31,115 dual-energy x-ray absorptiometry scans from UK Biobank, we identified 180 loci associated with seven highly heritable pelvic phenotypes. Birth canal phenotypes showed sex-specific genetic architecture, aligning with reproductive function. Larger birth canals were linked to slower walking pace and reduced back pain but increased hip osteoarthritis risk, whereas narrower birth canals were associated with reduced pelvic floor disorder risk but increased obstructed labor risk. Lastly, genetic correlation between birth canal and head widths provides evidence of coevolution between the human pelvis and brain, partially mitigating the dilemma.

12 November 2024 in Nature Communications
Verified
Journal Article
Open access
PDF file
DPDevansh Pandey
Devansh Pandey
MHMariana Harris
Mariana Harris
NGNandita R. Garud
Nandita R. Garud
VNVagheesh M Narasimhan
Vagheesh M Narasimhan
Large ancient DNA (aDNA) studies offer the chance to examine genomic changes over time, providing direct insights into human evolution. While recent studies have used time-stratified aDNA for selection scans, most focus on single-locus methods. We conducted a multi-locus genotype scan on 708 samples spanning 7000 years of European history. We show that the G12 statistic, originally designed for unphased diploid data, can effectively detect selection in aDNA processed to create ‘pseudo-haplotypes’. In simulations and at known positive control loci (e.g., lactase persistence), G12 outperforms the allele frequency-based selection statistic, SweepFinder2, previously used on aDNA. Applying our approach, we identified 14 candidate regions of selection across four time periods, with half the signals detectable only in the earliest period. Our findings suggest that selective events in European prehistory, including from the onset of animal domestication, have been obscured by neutral processes like genetic drift and demographic shifts such as admixture.

3 May 2024 in bioRxiv
Verified
Preprint
Open access
PDF file
LXLiaoyi Xu
Liaoyi Xu
EKEucharist Kun
Eucharist Kun
DPDevansh Pandey
Devansh Pandey
JWJoyce Y. Wang
Joyce Y. Wang
MBMarianne Brasil
Marianne Brasil
VNVagheesh M Narasimhan
Vagheesh M Narasimhan
Human pelvic shape has undergone significant evolutionary change since the divergence from the chimpanzee lineage. This transformation, involving the reduction of the pelvic canal size to support bipedal locomotion, is thought to give rise to the obstetrical dilemma, a hypothesis highlighting the mismatch between the large brain size of infants and the narrowed birth canal in females. Empirical evidence for this classic hypothesis has been equivocal, largely due to a lack of sample size and appropriate types of data. To elucidate the genetic underpinnings of pelvic morphology, we applied a deep learning model to 31,115 dual-energy X-ray absorptiometry (DXA) from the UK Biobank, extracting a set of seven pelvic proportion (PP) phenotypes, including measures of the birth canal. All PPs were found to be highly heritable (∼25-40%) and a genome-wide association study of these traits identified 179 independent loci. Unlike other skeletal proportions including long bone lengths, the subpubic angle associated with the birth canal exhibits a genetic correlation between sexes significantly less than 1, in line with sex-specific reproductive function. PPs were also left-right asymmetric but not heritable and instead associated with handedness. We conducted phenotypic and genetic association analyses to link PPs to 3 facets of the dilemma: locomotion, pelvic floor function and childbirth. Larger birth canal phenotypes were associated with reduced walking pace, decreased risk of back pain, and increased risk of hip osteoarthritis - phenotypes linked to locomotor efficiency. We also observed that a narrower birth canal width was associated with a reduced risk of pelvic floor disorders. When examining childbirth-related outcomes, narrower birth canal phenotypes were associated with increased risk of emergency cesarean sections and obstructed labor due to insufficient dilation, but not obstructed labor due to positioning of the fetus. Finally, we examined whether the dilemma might have been alleviated through evolution. We found no association between any PPs and gestational duration, contrary to the initial prediction by Washburn in 1960. However, we found that the birth weight of the child, a proxy for skull and brain size, was genetically correlated with birth canal width but not with other PPs. Collectively, our study offers fresh insight on a 60-year-old debate in human evolutionary studies. Our results support the idea that the obstetrical dilemma has played a central role in the co-evolution of the human brain and pelvis, while also highlighting the potential role of associated factors such as pelvic floor health.

25 April 2023 in bioRxiv
Verified
Preprint
Open access
PDF file
DPDevansh Pandey
Devansh Pandey
MHMariana Harris
Mariana Harris
NGNandita R. Garud
Nandita R. Garud
VNVagheesh M Narasimhan
Vagheesh M Narasimhan
Ancient DNA (aDNA) has been a revolutionary technology in understanding human history but has not been used extensively to study natural selection as large sample sizes to study allele frequency changes over time have thus far not been available. Here, we examined a time transect of 708 published samples over the past 7,000 years of European history using multi-locus genotype-based selection scans. As aDNA data is affected by high missingness, ascertainment bias, DNA damage, random allele calling, and is unphased, we first validated our selection scan, G12 ancient , on simulated data resembling aDNA under a demographic model that captures broad features of the allele frequency spectrum of European genomes as well as positive controls that have been previously identified and functionally validated in modern European datasets on data from ancient individuals from time periods very close to the present time. We then applied our statistic to the aDNA time transect to detect and resolve the timing of natural selection occurring genome wide and found several candidates of selection across the different time periods that had not been picked up by selection scans using single SNP allele frequency approaches. In addition, enrichment analysis discovered multiple categories of complex traits that might be under adaptation across these periods. Our results demonstrate the utility of applying different types of selection scans to aDNA to uncover putative selection signals at loci in the ancient past that might have been masked in modern samples.

10 October 2025 in medRxiv
Verified
Preprint
Open access
PDF file
DPDevansh Pandey
Devansh Pandey
LXLiaoyi Xu
Liaoyi Xu
EKEucharist Kun
Eucharist Kun
CLChenfei Li
Chenfei Li
JWJoyce Y. Wang
Joyce Y. Wang
VNVagheesh M Narasimhan
Vagheesh M Narasimhan
Coronary artery disease (CAD) is the leading cause of death worldwide, yet it is highly preventable. Early detection is critical, particularly because the first clinical manifestation of CAD is a heart attack in ∼50% of individuals. Current clinical risk scores rely largely on traditional biomarkers and do not leverage recent advances in medical imaging and genetics. To address this, we developed preCog , a multimodal artificial intelligence framework that integrates multi-organ imaging, genetic risk, metabolic biomarkers, ECGs and demographic data to predict time to incident CAD over up to 10 years of follow-up in a cohort of 60,000 UK Biobank participants. To quantify imaging risk, we applied 2D and 3D foundational computer vision models to more than 500,000 images spanning cardiac, liver and pancreas MRI, as well as DXA scans, each capturing distinct facets of CAD pathophysiology. We extracted deep learning derived image embeddings and compressed them to compact representations that were highly correlated with conventional imaging metrics of cardiac function (e.g. ejection fraction and stroke volumes) yet outperformed these metrics for CAD prediction (AUC 0.794 vs 0.666). Because imaging cost is a key constraint, we evaluated modality-level contributions and found that only cardiac long-axis cine MRI and aortic distensibility MRI contributed substantial independent value. After adjusting for baseline traits, liver, pancreas and DXA features added no significant predictive power. We constructed a joint time-to-event model integrating imaging, a polygenic risk score (PRS) trained on more than 1.25 million individuals from non-imaged UK Biobank participants, the Million Veteran Program and FinnGen, blood biochemistry and clinical variables. The joint model achieved a C-index of 0.75, exceeding PREVENT (0.71) and Framingham Risk Score (0.66). Importantly we found that imaging and genetic risks were largely independent, indicating that individuals with similar imaging-based risk at a given age may progress differently based on underlying genetic risk. A hierarchical risk stratification framework combining clinical, genetic and imaging data identified a subgroup with a 15-fold increase in incident CAD risk relative to a low-risk baseline. Performance was consistent across data collection centers spanning rural and urban settings and diverse demographics (C-index range 0.73–0.76). Our findings demonstrate the utility of multi-modal AI for medical forecasting of common complex disease to preempt or mitigate their occurrence.

4 September 2026 in medRxiv
Verified
Preprint
Open access
PDF file
DPDevansh Pandey
Devansh Pandey
VNVagheesh M Narasimhan
Vagheesh M Narasimhan
Self-supervised models increasingly convert medical images into quantitative phenotypes for biological discovery, but statistical reproducibility does not establish that a learned phenotype represents the intended anatomy. We trained a video masked-autoencoder on 69,932 UK Biobank cardiac cine-MRI studies and performed genome-wide association analysis of its latent representation. Although 18 of 20 leading axes were heritable with well-calibrated statistics, the representation encoded substantial field-of-view information: body size, stature and imaging centre (linear-probe R 2 = 0.55 for site); standard genomic-control and LD-score diagnostics did not identify this source of phenotype-level confounding. Restricting the field of view to the heart and residualising body and acquisition covariates before dimensionality reduction substantially attenuated linear and non-linear nuisance information while retaining cardiac signal. Adjusting the same covariates only during association testing attenuated nuisance associations but recovered substantially less of the cardiac-associated genetic signal, consistent with nuisance variation having already influenced the principal-component basis. The corrected representation identified new associated loci beyond those detected using supervised phenotypes at matched sample size, which shared genetic architecture selectively with cardiac-conduction traits and were localised to cardiac structures within the imaged field of view. Confounding in learned medical phenotypes can arise upstream of association testing, highlighting the importance of auditing and, where appropriate, correcting learned representations before association testing.