Akhil Kumar

I am an Associate Computational Researcher in Tsankov Lab at the Icahn School of Medicine at Mount Sinai, where I integrate bulk, single-cell, and spatial transcriptomics to stratify microsatellite-stable colorectal cancer. Before this, I was a Research Fellow in Perumal Lab at the School of Biological Sciences of Indian Institute of Technology (IIT) Delhi, where I used evolutionary bioinformatics approaches to better understand host adaptation dynamics of SARS-CoV-2 and Hepatitis C virus. I received my B.Tech. in Chemical Engineering from IIT Delhi.

Email / GitHub / Google Scholar / LinkedIn / CV

Published Research

I'm interested in addressing biomedical problems, particularly in cancer, through functional genomics and systems biology, with an emphasis on computational method development. Selected works are highlighted.

	A cellular and spatial atlas of TP53-associated tissue remodeling in lung adenocarcinoma. William Zhao, Thinh T Nguyen, Atharva Bhagwat, Akhil Kumar, Bruno Giotti, ..., Aaron N Hata , Aviv Regev , Bruce E Johnson , Alexander M Tsankov Nature Cancer*, 2025 paper / code / We constructed a multiomic cellular and spatial atlas of 23 treatment-naive human lung tumors to define how TP53 mutations affect the lung adenocarcinoma tumor microenvironment. I validated the trends observed in single-cell data in two independent bulk cohorts using regression models, survival analysis, and transcriptome-based estimates of differentiation potency, and documented my analysis pipelines for public release.
	Cancer Cell of Origin. Mohamad D Bairakdar, Wooseung Lee, Bruno Giotti, Akhil Kumar, Paula Stancl, Elvin Wagenblast, Dolores Hambardzumyan, Paz Polak, Rosa Karlic, Alexander M Tsankov Nature Communications*, 2025 paper / code / We combined 3,669 whole genome sequencing patient samples, 559 single-cell chromatin accessibility cellular profiles, and machine learning (ML) to predict the cell of origin of 37 cancer subtypes, confirming both the known anatomical and cellular origins of numerous cancers, often at cell subset resolution. The model pointed to an intermediate metaplastic state shared by several gastrointestinal cancers; I tested this by analyzing six single-cell RNA-seq datasets of injury- and oncogene-induced metaplasia, showing that metaplastic cells converge on gastric goblet cell programs.
	Highly conserved genomic regions in SARS-CoV-2 as diagnostic targets. Akhil Kumar, Rishika Kaushal, Himanshi Sharma, Khushboo Sharma, Manoj B Menon, Vivekanandan Perumal Briefings in Functional Genomics, 2023 paper / code / We identified 11 conserved stretches in over 6.3 million SARS-CoV-2 genomes including all the major variants of concerns. Each conserved stretch is ≥100 nucleotides in length with ≥99.9% conservation at each nucleotide position. Interestingly, six of the eight conserved stretches in ORF1ab overlapped significantly with well-folded experimentally verified RNA secondary structures. Our findings highlight the role of structural constraints at both RNA and protein levels that contribute to the sequence conservation of specific genomic regions in SARS-CoV-2.
	Host-adaptation dynamics in Hepatitis C virus. Sanket Mukherjee, Akhil Kumar, Jasmine Samal, Ekta Gupta, Vivekanandan Perumal, Manoj B Menon Pathogens, 2022 paper / code / This work highlights CpG depletion in HCV genomes during their evolution in humans and the role of ZAP-mediated selection in HCV evolution.
	Host-adaptation dynamics in SARS-CoV-2. Akhil Kumar, Nishank Goyal, Nandhini Saranathan, Sonam Dhamija, Saurabh Saraswat, Manoj B Menon, Vivekanandan Perumal Molecular Biology and Evolution, 2022 paper / code / This work highlights how temporal variations in selection pressures during virus adaption may impact the rate and the extent of CpG depletion in virus genomes.

Work in Progress

Ongoing projects and manuscripts in preparation.

Stratifying microsatellite-stable colorectal cancer by its tumor microenvironment.

Akhil Kumar, ..., Alexander M Tsankov
Ongoing, 2026

My primary project integrates bulk, single-cell, and spatial transcriptomics to define molecular subtypes of microsatellite-stable colorectal cancer and understand how their tumor microenvironments differ.

Early recurrence of mutations in SARS-CoV-2 predicts predominant mutations.

Akhil Kumar*, Rishika Kaushal*, Manoj B Menon, Vivekanandan Perumal
In preparation, 2026
code /

We called mutations in 15.5 million SARS-CoV-2 genomes from GISAID (5.6 million after quality filtering, December 2019 to December 2025) and found that early recurrence predicts which mutations become predominant: 79–85% of early mutations that later reached ≥1% frequency had recurred within 28 days of first detection.

Additional Research

A collection of prior projects.

	Temporal evolution of mono- and dinucleotides in human mitochondrial DNA Perumal Lab, Indian Institute of Technology Delhi Independent study 2019-11-28 slides / Curated and filtered 889 human mtDNA sequences spanning ~50,000 years from the Ancient mtDNA Database. Tracked shifts in mono- and dinucleotide composition and observed/expected ratios across time-binned samples, and found significant shifts in TT, AG, and CT dinucleotide ratios over time, with opposing purine trends on the two strands.
	In-silico design of small ligand molecules to inhibit early-stage insulin aggregation nucleation Multiscale Modeling Group, Indian Institute of Technology Delhi Bachelor's thesis 2019-11-24 slides / Contributed to the in silico design of small-molecule ligands to inhibit early-stage nucleation of insulin aggregation. Identified EGCG and polyoxometalates as candidate ligands through literature review and computational modeling, and docked EGCG into a partially folded insulin intermediate using ROSIE, the Rosetta modeling web server.
	Systematic construction and validation of kinetic model from metabolic networks Biomolecular Computational Group, Indian Institute of Science Bangalore Summer Internship 2019-07-15 slides / Compared kinetic rate laws (mass-action, Michaelis–Menten, and convenience kinetics) for building dynamic metabolic models. Built a convenience-kinetics model of glycolysis, the TCA cycle, and the respiratory chain in COPASI (92 reactions, 110 species, 4 compartments), simulated its dynamics, and ran parameter sensitivity analysis.
	Single-cell transcriptomic data analysis Srinivas Group, University of Oxford Summer Internship 2018-07-24 code / Processed raw Smart-seq2 data from wild-type and Aspp2 mutant early mouse embryos through Salmon quantification. Applied six-metric cell quality control (498 high-quality cells retained), scran normalization, and highly variable gene selection, then resolved epiblast, mesoderm, visceral endoderm, and anterior visceral endoderm populations by clustering and marker genes. NOTE: The code repository for this project is private, please contact me for access.