About

Manolis Kellis is a Professor of Computer Science at MIT, specializing in Artificial Intelligence, Machine Learning, and their applications to healthcare and life sciences. His research areas include AI for Healthcare and Life Sciences, Biological and Medical Devices and Systems, and AI + Decision-making, combining traditions from computer science and electrical engineering to develop techniques for systems that interact with the external world through perception, communication, and action, while also learning, making decisions, and adapting to changing environments. His work involves leveraging computational, theoretical, and experimental tools to develop groundbreaking sensors, energy transducers, physical substrates for computation, and systems addressing shared human challenges. Kellis's research has contributed to understanding gene expression in Alzheimer’s disease, improving genetic prediction models across diverse populations, and transforming micronutrient dosing to improve child health in Nigeria.

Research topics

• Biology
• Genetics
• Computational biology
• Evolutionary biology
• Medicine
• Computer Science
• Internal medicine
• Cell biology
• Psychology
• Pathology
• Neuroscience
• Mathematics
• Statistics
• Chemistry
• Ecology
• Bioinformatics
• Psychiatry
• Endocrinology

Selected publications

• Epigenetically constrained astrocyte states underlie prefrontal cortex vulnerability in Down syndrome–associated Alzheimer’s disease

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-04-21

  articleOpen accessCorresponding

  Down syndrome (DS), caused by trisomy 21, confers a near-universal risk for Alzheimer's disease (AD), yet individuals exhibit marked variability in cognitive decline, suggesting the presence of cellular mechanisms that modulate vulnerability and resilience. However, these mechanisms remain poorly defined in the human brain. Here, we integrate matched single-nucleus RNA-seq and ATAC-seq profiles from the prefrontal cortex (PFC) and amygdala (AMY) of age-matched individuals with DS with and without AD (DSAD), enabling direct comparison within a shared genetic background. We identify basal astrocytes in the PFC as a selectively vulnerable cell state in DSAD, characterized by both reduced abundance and coordinated transcriptional and regulatory reprogramming. This state exhibits a shift away from homeostatic support functions, with decreased cytokine signaling and lipid-handling programs, alongside increased steroid- and nuclear receptor-associated activity. Concomitantly, chromatin accessibility profiling reveals reduced engagement of immune- and stress-responsive transcription factor programs, including AP-1, STAT, and BACH families, with linked regulatory perturbations at loci such as ABCA1, DAB2IP, and IL1RAP. Together, these findings define a previously unrecognized astrocyte state marked by epigenetic constraint and diminished responsiveness to stress and inflammatory signals, distinguishing it from classical reactive astrocyte phenotypes. Our results nominate PFC basal astrocytes as a key locus of vulnerability in DSAD and suggest that failure to mount appropriate astrocyte responses, rather than overt activation alone, may contribute to neurodegenerative progression.

  PublisherDOI

• Expansion Revealing of Pathology Resolves Nanostructures Associated with Inflammatory Phenotypes in COVID-19 Decedent Human Brain Tissue

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-05-15

  articleOpen access

  Abstract Expansion revealing (ExR) elucidates cellular organization by separating proteins within dense nanostructures by 20x linear expansion, but requires fixation procedures incompatible with human pathology specimens. Here, we report ExR of pathology (ExRPath), which attains ∼20 nm resolution and decrowding of such tissues, through iterative 20x expansion, adapted to human brain pathology specimens. We also report a single-shot 15x expansion protocol for such tissues (15ExMPath), achieved through one-shot 15x expansion. Applying ExRPath and 15ExMPath to COVID-19-decedent brain tissue reveals periodic amyloid nanoclusters that co-localize with SARS-CoV-2 in a rare minority of patient specimens, pointing to a potential neuroinflammatory phenotype associated with COVID-19, and highlighting the power of high-throughput nanoimaging, empowered by expansion microscopy, for discovering potential novel disease mechanisms.

  PublisherOA PDFDOI

• Role of AI in drug development: Current status, challenges, opportunities, and future promise

  Artificial Intelligence in Health · 2026-03-13

  articleOpen accessSenior author

  Artificial intelligence (AI) heralds a transformative shift in drug development, with speed, precision, and predictive power as its core features. Advances in systems-level biology platforms, coupled with substantial investments in generative AI-centric pharma integration, have fostered healthy optimism among stakeholders about identifying new cures through renewed approaches and improved productivity. However, navigating epistemological, ethical, patient safety, and ontological dimensions within research and development (R&D) presents challenges that AI must address to enhance its mainstream adoption and practical utility. Here, multidisciplinary experts discuss key applications of AI across the full continuum of drug development, examine the challenges encountered, and propose solution frameworks. Drug development remains fraught with unknown biology, patient heterogeneity, and perplexing therapeutic risks. Stringent regulatory and compliance guidelines further necessitate that conventional pharma processes, practices, and strategies remain paramount in R&D execution, while guiding the integration of AI in a “value-for-effort,” evidence-based, yet Promethean fashion.

  PublisherOA PDFDOI

• Dissecting Alzheimer’s disease heterogeneity by cross-trait polygenic prediction

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-05-15

  articleOpen access

  Abstract Mapping the genetic basis of inter-individual heterogeneity in multifactorial diseases opens the door to mechanistic insights and opportunities for targeted intervention. In Alzheimer’s disease (AD), clinical and pathological heterogeneity is well recognized, but genetic dissection is limited by a lack of well-powered cohorts with deep phenotypic characterization. Here, we introduce a polygenic score (PGS) analysis strategy to address these limitations by leveraging the inherent pleiotropy in complex trait genetics. We perform a cross-cohort, cross-trait application of pre-trained PGS, integrating 713 UK Biobank-derived PGS with 36 deep AD phenotypes across 1678 ROSMAP participants. We identify 268 statistically significant (FDR<0.1) associations between 12 prioritized PGS and 36 AD phenotypes. Prioritized PGS include blood lipid measurements, inflammatory biomarkers, and cancer traits; observed AD phenotypes include cognition, amyloid, and tangles. Of the 268 associations, 49 persist with APOE -excluded PGS. Predictive models trained on multiple prioritized PGS outperform the AD PGS or APOE alone for predicting amyloid and cognition. Lastly, our approach identifies six individual-level AD polygenic subtypes supported by distinct pathological patterns. Overall, we combine large-scale biobank resources and deeply-phenotyped cohorts using PGS, reveal genetic features underlying AD heterogeneity, and provide a general model for stratifying heterogeneous disease-focused cohorts using genomics.

  PublisherDOI

• Intraspecific sequence variation and complete genomes refine the identification of rapidly evolved regions in humans

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-10-21 · 3 citations

  preprintOpen access

  Summary Humans exhibit significant phenotypic differences from other great apes, yet pinpointing the underlying genetic changes has been limited by incomplete reference genomes and a reliance on single assemblies to represent a species. We aligned 20 telomere-to-telomere (T2T) assemblies spanning great ape divergence and variation to define 1,596 Consensus HAQERs (consensus human ancestor quickly evolved regions), regions that diverged rapidly between the human-chimpanzee ancestor and an ancestral node of modern humans. Unlike prior HAQER sets, Consensus HAQERs incorporate population variation, reducing the likelihood of intraspecies variation appearing as interspecies divergence. Consensus HAQERs exhibit signatures of elevated mutation rates, ancient positive selection, bivalent regulatory function, are enriched in disease-linked loci, and often emerged in previously inaccessible repetitive DNA. Through multiplex, single-cell enhancer assays, we identify HAQERs as active enhancers in the developing brain and cardiomyocytes, highlighting their potential contributions to human-specific gene regulation across multiple tissues. Highlights ● Telomere-to-telomere alignments of diverse human and great ape genomes identify 1,596 Consensus HAQERs, regions of rapid sequence divergence separating human ancestors from other great apes. ● Consensus HAQERs exhibit signatures of both elevated mutation rates and ancient positive selection. ● HAQERs are enriched in bivalent regulatory elements and disease-linked loci. ● Multiplex, single-cell gene regulatory assays identify HAQERs as enhancers in the developing heart and brain.

  PublisherDOI

• CONCERT predicts niche-aware perturbation responses in spatial transcriptomics

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-11-10 · 1 citations

  preprintOpen access

  Spatial perturbation transcriptomics measures how genetic or chemical edits alter gene expression while preserving tissue context. Perturbation outcomes depend on a cell's intrinsic state and also on how effects propagate across cellular microenvironments. We present CONCERT, a niche-aware generative model that embeds perturbation context and learns spatial kernels with a Gaussian process variational autoencoder to predict perturbation effects across tissue. We formalize three tasks: patch, border, and niche, predicting responses in nearby unperturbed regions, at tissue interfaces, and as a function of surrounding microenvironments. We evaluate CONCERT on Perturb-map lung datasets. CONCERT outperforms state-of-the-art models (dissociated counterfactuals, spatialized perturbation models, and kNN), reducing E-distance by up to 33.77% (patch), 26.05% (border), and 33.74% (niche) versus the next best, with mean absolute error down by up to 23.28% and Pearson correlation up by up to 9.10%. Two case studies go beyond benchmarking. In dextran sodium sulfate-induced colitis, CONCERT reconstructs spatial gene expression at unmeasured time points, produces longitudinal comparisons across unpaired mice, resolves inter-mouse heterogeneity, and recovers consistent temporal declines of inflammation-associated genes across regions. In ischemic stroke, CONCERT predicts responses under variable lesion sizes and in a 3D formulation across brain sections, capturing lesion-core and peri-lesion patterns. CONCERT performs niche-aware counterfactual prediction, reconstructs missing spatial data, and models perturbation responses across tissues.

  PublisherOA PDFDOI

• sc4D: spatio-temporal single-cell transcriptomics analysis through embedded optimal transport identifies joint glial response to Alzheimer’s disease pathology

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-11-19

  preprint

  Abstract A precise understanding of disease-associated spatio-temporal transcriptional dynamics is critical for nominating therapeutic interventions that drive desired perturbation responses. In disease contexts, pathological processes unfold across diverse cellular states, spatial environments, and timescales; however, current computational approaches have a limited ability to jointly model these complex dynamics or infer cellular trajectories from in silico perturbation experiments. Here, we present sc4D, a biologically interpretable spatio-temporal (4D) analysis framework for single-cell transcriptomics in disease studies, integrating autoencoder embeddings with optimal transport. Applying sc4D to longitudinal spatial transcriptomics samples from Alzheimer’s disease mouse models, we report known disease biology and novel, testable mechanisms, including late-stage microglia-astrocyte syncytium near amyloid-β plaques. In silico perturbations predict the restoration of protective, anti-inflammatory microglia through CAMTA1 activation or Donepezil, as validated by independent experimental findings. Overall, our work highlights the critical benefits of joint spatio-temporal modeling for elucidating disease mechanisms and predicting candidate interventions to improve cellular response.

  PublisherDOI

• Disparities and trends in global representation of human genetics conferences: a 26-year longitudinal study of ASHG and ESHG

  medRxiv · 2025-08-16 · 1 citations

  preprintOpen access

  Abstract Equity in human genetics research requires balanced participation not only from study participants from global populations but also from the researchers who drive the science. While disparities among research participants across ancestries and countries have been well studied, the representation and disparities of researchers themselves on the global stage remains poorly understood. Here, we analyzed over 100,000 abstracts presented at two leading annual conferences in the field, the American Society of Human Genetics (ASHG) and the European Society of Human Genetics (ESHG), from 1999 to 2024 to assess trends and geographic disparities. North America and Europe consistently dominated abstract contributions, whereas continents such as Africa, Oceania, and East Asia remained underrepresented, despite gradual increases in participation. The imbalance was even more pronounced in oral presentation: at ASHG, abstracts from North America were approximately 4 times more likely to be selected for talks than those from East Asia and 23 times more likely than those from South America; at ESHG, Europe’s advantage was 2 times and 9 times, respectively. Notably, Oceania had the highest relative success in oral presentation, with a ratio 5 times higher than East Asia and 29 times higher than South America in ASHG, and 8 times and 33 times higher, respectively, in ESHG. To explore potential drivers of these disparities, we examined 6 national level variables. The multivariable regression model indicated that GDP is the primary factor for abstract, while Nature Index Share is the main factor for oral presentation counts. Our findings highlight persistent global inequalities in representation of human geneticists at premier conferences. Greater international support and targeted initiatives are needed to promote more equitable worldwide involvement in human genetics.

  PublisherOA PDFDOI

• Multi-omics analysis of a pig-to-human decedent kidney xenotransplant

  Nature · 2025-11-13 · 13 citations

  articleOpen access
  PublisherOA PDFDOI

• Type 2 Diabetes and Obesity Alter Exercise Training-Induced Transcriptional Adaptations to Subcutaneous White Adipose Tissue

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-10-12

  preprintOpen access

  Abstract White adipose tissue (WAT) dysfunction contributes to obesity-associated metabolic disease and type 2 diabetes (T2D). Rodent studies have demonstrated that exercise training improves WAT function, but molecular studies investigating exercise training effects on WAT in humans have been limited, particularly in the context of metabolic disease. Here, we defined the subcutaneous WAT (scWAT) transcriptome in middle-aged adults (10 male, 19 female) that were classified by lower BMI (<27 kg/m 2 ), higher BMI (≥27 kg/m 2 ), and T2D status before and after a 10-week endurance exercise regimen. At baseline, 624 genes were significantly upregulated and 112 genes downregulated in the scWAT from higher BMI participants compared to lower BMI. There was a spectrum of pathway dysregulation in scWAT in higher BMI individuals, ranging from increased markers of extracellular matrix (ECM) deposition and inflammation to decreased circadian rhythm gene expression. In people with T2D, there were additional transcriptomic differences such as translation-related pathways, selenoamino acid metabolism, and proteoglycans. Exercise training had robust effects on the transcriptome regardless of metabolic status, and notably, for the high BMI and T2D groups, training reversed several of the detrimental gene expression patterns in a cell-type-specific manner. These beneficial exercise-induced transcriptomic adaptations significantly correlated with lower levels of free fatty acids and blood pressure, particularly in participants with higher BMI and T2D. By integrating our exercise training-modulated genes with GWAS meta-analysis of physical activity, genes influenced by exercise training in the higher BMI group showed a significant enrichment in genetic associations of exercise traits in the population. A circadian rhythm-related transcription factor NR1D1 was enriched in enhancers linked with both the exercise differentially expressed genes (DEGs) and GWAS signals, suggesting a link between the circadian rhythm and training-induced adaptations. These findings demonstrate that obesity and T2D result in marked, progressive alterations in cell-type specific gene transcription in scWAT, while endurance exercise training reverses many of the metabolic disease-associated adaptations. Identification of novel molecular pathways regulated by exercise training can lead to therapeutic targets for obesity and metabolic disease.

  PublisherOA PDFDOI

Recent grants

• NIH Grant U01HG007610

  NIH · $1.7M · 2018

• Network-based prediction and validation of causal schizophrenia genes and variants

  NIH · $2.1M · 2018–2021

• Single-cell transcriptional and epigenomic dissection of Alzheimer's Disease and Related Dementias

  NIH · $4.0M · 2018–2022

• NIH Grant R01HG008155

  NIH · $2.3M · 2019

• NIH Grant RC2HG005639

  NIH · $4.1M · 2013

Frequent coauthors

• Irwin Jungreis

  Massachusetts Institute of Technology

  126 shared

• Li‐Huei Tsai

  96 shared

• Rory Johnson

  University Hospital of Bern

  81 shared

• Eloi Schmauch

  University of Eastern Finland

  76 shared

• Anshul Kundaje

  Stanford University

  75 shared

• Jason Ernst

  University of California, Los Angeles

  74 shared

• Mark Gerstein

  72 shared

• José Dávila-Velderrain

  70 shared

Labs

• MIT EECS - Manolis Kellis LabPI