About

Madhav Mani is an Associate Professor of Engineering Sciences and Applied Mathematics at Northwestern University. His research focuses on developing novel mathematical abstractions to provide insights into biological phenomena and data. He pursues lines of inquiry that attempt to deepen understanding of life through the lens of mathematics, working on diverse biological topics such as organismal development, cellular physiology and structure, and ecological dynamics. His approach involves close collaboration with modern biological data, utilizing data-driven and AI methods alongside mathematical and theoretical physics principles, within long-term partnerships with experimental biology groups. Mani's educational background includes a Ph.D. in Applied Mathematics and an S.M. in Engineering Sciences from Harvard University, as well as advanced studies in Mathematics and Theoretical Physics at Cambridge University. His scientific vision emphasizes inspiring young scientists to enjoy the pursuit of mathematics and science, fostering their talents and tastes. His contributions have been recognized through awards such as the Simons Postdoctoral Fellowship, the Derek Bok Undergraduate Teaching Award, and the Robert L. Wallace Prize Fellowship.

Research topics

• Computer Science
• Artificial Intelligence
• Biology
• Computational biology
• Bioinformatics
• Genetics
• Evolutionary biology
• Ecology
• Biochemistry

Selected publications

• Learning the Shape of Evolutionary Landscapes: Geometric Deep Learning Reveals Hidden Structure in Phenotype-to-Fitness Maps

  PRX Life · 2026-01-16 · 1 citations

  articleOpen access

  Elucidating the complex relationships between genotypes, phenotypes, and fitness remains one of the fundamental challenges in evolutionary biology. Part of the difficulty arises from the enormous number of possible genotypes and the lack of understanding of the underlying phenotypic differences driving adaptation. Here we present a computational method that takes advantage of modern high-throughput fitness measurements to learn a map from high-dimensional fitness profiles to a low-dimensional latent space in a geometry-informed manner. We demonstrate that our approach using a Riemannian Hamiltonian variational autoencoder outperforms traditional linear dimensionality reduction techniques by capturing the nonlinear structure of the phenotype-fitness map. When applied to simulated adaptive dynamics, we show that the learned latent space retains information about the underlying adaptive phenotypic space and accurately reconstructs complex fitness landscapes. We then apply this method to a dataset of high-throughput fitness measurements of under different antibiotic pressures and demonstrate superior predictive power for out-of-sample data compared to linear approaches. Our work provides a data-driven implementation of Fisher's geometric model of adaptation, transforming it from a theoretical framework into an empirically grounded approach for understanding evolutionary dynamics using modern deep learning methods.

  PublisherDOI

• Data from: A phase oscillator model of cell cycles reveals nuclear density control in a branched fungal network

  Research Data Repository, Duke University · 2026-01-01

  datasetOpen access

  Microscopy data from the paper: Data from: A phase oscillator model of cell cycles reveals nuclear density control in a branched fungal network. Paper abstract: Maintaining an appropriate nuclear-to-cytoplasmic ratio is essential across cell types for physiological function, and mechanisms of size control have been extensively studied in mononucleate cells. Much less is known about how comparable control is achieved in cells where many nuclei share a common cytoplasm, which are seen in many contexts including muscle, placenta and filamentous fungi. The filamentous fungus Ashbya gossypii forms a branching mycelial network in which individual nuclei divide asynchronously, while the number of nuclei per cell volume (the nuclear density) is tightly controlled. How global regulation of nuclear density coexists with local cell cycle asynchrony remains unclear. To address this we model nuclei as a dividing population of phase oscillators within a branching cell network and parameterize the model with measurements from Ashbya cells. The model demonstrates that asynchrony is required to prevent large density fluctuations that would result from synchronous division, and that introducing a nuclear-density checkpoint to the cell cycles leads to synchrony if it is the only mechanism of density control. We find that coupling branch formation to nuclear density both stabilizes nuclear density and prevents the emergence of synchronous cycles. Supporting these predictions, we demonstrate that mutants with branching defects and increased cell cycle synchrony display greater variability in nuclear density. Our results indicate that asynchronous nuclear cycles together with density-responsive branching maintain a constant nuclear density, revealing a strategy for regulating the nuclear-to-cytoplasmic ratio in large multinucleate cells.

  PublisherDOI

• Illuminating the uncharacterized regulatory genome of <i>E. coli</i> with massively parallel reporters

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-05-15

  preprintOpen access

  1 Abstract All cells respond to changes in both their internal milieu and the environment around them through the regulation of their genes. Despite decades of effort, there remain huge gaps in our knowledge of both the function of many genes (the so-called y-ome) and how they adapt to changing environments via regulation. Here we describe a joint experimental and theoretical dissection of the regulation of a broad array of more than 100 biologically interesting genes in E. coli across 39 diverse environments, enabling us to identify the binding sites and transcription factors that mediate regulatory control. Using a combination of mutagenesis, massively parallel reporter assays, mass spectrometry, and tools from information theory and statistical physics, we go from complete ignorance of a promoter’s environment-dependent regulatory architecture to a quantitative description of its binding sites, candidate transcription factors that bind them where identifiable, and the conditions under which they act. As proof of principle of the biological insights to be gained from such a study, we chose a combination of genes from the y-ome, toxin-antitoxin pairs, and genes hypothesized to be part of regulatory modules; we discovered a host of new insights into their underlying regulatory landscape and resulting biological function. We highlight discoveries for y-ome genes, including transcription start sites and transcription factor binding sites at base-pair resolution, and their dependence on growth conditions.

  PublisherOA PDFDOI

• Author response: Spectral decomposition unlocks ascidian morphogenesis

  2025-06-16

  peer-reviewOpen accessSenior author
  PublisherDOI

• Learning the Shape of Evolutionary Landscapes: Geometric Deep Learning Reveals Hidden Structure in Phenotype-to-Fitness Maps

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-05-07 · 2 citations

  preprintOpen access

  Elucidating the complex relationships between genotypes, phenotypes, and fitness remains one of the fundamental challenges in evolutionary biology. Part of the difficulty arises from the enormous number of possible genotypes and the lack of understanding of the underlying phenotypic differences driving adaptation. Here, we present a computational method that takes advantage of modern high-throughput fitness measurements to learn a map from high-dimensional fitness profiles to a low-dimensional latent space in a geometry-informed manner. We demonstrate that our approach using a Riemannian Hamiltonian Variational Autoencoder (RHVAE) outperforms traditional linear dimensionality reduction techniques by capturing the nonlinear structure of the phenotype-fitness map. When applied to simulated adaptive dynamics, we show that the learned latent space retains information about the underlying adaptive phenotypic space and accurately reconstructs complex fitness landscapes. We then apply this method to a dataset of high-throughput fitness measurements of E. coli under different antibiotic pressures and demonstrate superior predictive power for out-of-sample data compared to linear approaches. Our work provides a data-driven implementation of Fisher’s geometric model of adaptation, transforming it from a theoretical framework into an empirically grounded approach for understanding evolutionary dynamics using modern deep learning methods.

  PublisherOA PDFDOI

• Functional regimes define soil microbiome response to environmental change

  Nature · 2025-07-16 · 29 citations

  articleOpen access

  Abstract The metabolic activity of soil microbiomes has a central role in global nutrient cycles 1 . Understanding how soil metabolic activity responds to climate-driven environmental perturbations is a key challenge 2,3 . However, the ecological, spatial and chemical complexity of soils 4–6 impedes understanding how these communities respond to perturbations. Here we address this complexity by combining dynamic measurements of respiratory nitrate metabolism 7 with modelling to reveal functional regimes that define soil responses to environmental change. Measurements across more than 1,500 soil microcosms subjected to pH perturbations 8,9 reveal regimes in which distinct mechanisms govern metabolite dynamics. A minimal model with two parameters, biomass activity and growth-limiting nutrient availability, predicts nitrate utilization dynamics across soils and pH perturbations. Parameter shifts under perturbation reveal three functional regimes, each linked to distinct mechanisms: (1) an acidic regime marked by cell death and suppressed metabolism; (2) a nutrient-limited regime in which dominant taxa exploit matrix-released nutrients; and (3) a resurgent growth regime driven by exponential growth of rare taxa in nutrient-rich conditions. We validated these model-derived mechanisms with nutrient measurements, amendment experiments, sequencing and isolate studies. Additional experiments and meta-analyses suggest that functional regimes are widespread in pH-perturbed soils.

  PublisherOA PDFDOI

• Author response: Spectral decomposition unlocks ascidian morphogenesis

  2025-04-15

  peer-reviewOpen accessSenior author

  Describing morphogenesis generally consists in aggregating the multiple high resolution spatiotemporal processes involved into reproducible low dimensional morphological processes consistent across individuals of the same species or group. In order to achieve this goal, biologists often have to submit movies issued from live imaging of developing embryos either to a qualitative analysis or to basic statistical analysis. These approaches, however, present noticeable drawbacks, as they can be time consuming, hence unfit for scale, and often lack standardisation and a firm foundation. In this work, we leverage the power of a continuum mechanics approach and flexibility of spectral decompositions to propose a standardised framework for automatic detection and timing of morphological processes. First, we quantify whole-embryo scale shape changes in developing ascidian embryos by statistically estimating the strain-rate tensor field of its time-evolving surface without the requirement of cellular segmentation and tracking. We then apply to this data spectral decomposition in space using spherical harmonics and in time using wavelets transforms. These transformations result in the identification of the principal dynamical modes of ascidian embryogenesis and the automatic unveiling of its blueprint in the form of scalograms that tell the story of development in ascidian embryos.

  PublisherDOI

• Spectral decomposition unlocks ascidian morphogenesis

  eLife · 2025-05-19

  articleOpen accessSenior author

  Describing morphogenesis generally consists in aggregating the multiple high-resolution spatiotemporal processes involved into reproducible low-dimensional morphological processes consistent across individuals of the same species or group. In order to achieve this goal, biologists often have to submit movies issued from live imaging of developing embryos either to a qualitative analysis or to basic statistical analysis. These approaches, however, present noticeable drawbacks as they can be time consuming, hence unfit for scale, and often lack standardization and a firm foundation. In this work, we leverage the power of a continuum mechanics approach and flexibility of spectral decompositions to propose a standardized framework for automatic detection and timing of morphological processes. First, we quantify whole-embryo scale shape changes in developing ascidian embryos by statistically estimating the strain rate tensor field of its time-evolving surface without the requirement of cellular segmentation and tracking. We then apply to this data spectral decomposition in space using spherical harmonics and in time using wavelets transforms. These transformations result in the identification of the principal dynamical modes of ascidian embryogenesis and the automatic unveiling of its blueprint in the form of scalograms that tell the story of development in ascidian embryos.

  PublisherDOI

• Stomatal setpoints and environmental responsiveness are sculpted by developmental trajectories

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-12-25 · 1 citations

  articleOpen access

  SUMMARY Efficient gas and water exchange between plants and their environment largely depends on the number and distribution of stomata, cellular valves in leaf epidermis. Core genetic regulators of stomatal cell identity and pattern along with asymmetric stem-cell like divisions in stomatal precursors are hypothesized to customize stomatal production for optimal leaf performance. How these regulators work in concert and how division dynamics are modified and adjusted in different environments, however, are poorly understood. Here, we leveraged the variation in stomatal patterning in Arabidopsis thaliana accessions from diverse environments to define developmental rules and constraints in the stomatal lineage. The accessions’ subtle and quantitative variation enables us to identify which cellular parameters are flexible, revealing how developmental plasticity generates phenotypic plasticity. By developing live-cell imaging tools to track cellular behaviors during leaf growth under varying environmental conditions in these accessions, we could decompose stomatal density variation into its developmental origins. Variation in final stomatal numbers is driven by differences in the relative contributions of stomatal initiation, cell size-based fate thresholds, general proliferative capacity, and coordination between sister and neighbor cell behaviors. Overall, diverse accessions converge toward two lineage regimes: one dominated by autonomous decisions with loose cell-cell coordination, the other by extensive cell-cell coordination. Challenging accessions with environmental fluctuations revealed regime-specific flexibility, with plasticity primarily mediated by a single division-related parameter. Our results show how cellular parameters integrate into alternative developmental strategies that shape environmental responsiveness.

  PublisherOA PDFDOI

• Spectral decomposition unlocks ascidian morphogenesis

  eLife · 2025-04-15

  preprintOpen accessSenior author

  Abstract Describing morphogenesis generally consists in aggregating the multiple high resolution spatiotemporal processes involved into reproducible low dimensional morphological processes consistent across individuals of the same species or group. In order to achieve this goal, biologists often have to submit movies issued from live imaging of developing embryos either to a qualitative analysis or to basic statistical analysis. These approaches, however, present noticeable drawbacks, as they can be time consuming, hence unfit for scale, and often lack standardisation and a firm foundation. In this work, we leverage the power of a continuum mechanics approach and flexibility of spectral decompositions to propose a standardised framework for automatic detection and timing of morphological processes. First, we quantify whole-embryo scale shape changes in developing ascidian embryos by statistically estimating the strain-rate tensor field of its time-evolving surface without the requirement of cellular segmentation and tracking. We then apply to this data spectral decomposition in space using spherical harmonics and in time using wavelets transforms. These transformations result in the identification of the principal dynamical modes of ascidian embryogenesis and the automatic unveiling of its blueprint in the form of scalograms that tell the story of development in ascidian embryos.

  PublisherDOI

Recent grants

• Collaborative Research: MTM 1: Decoding the genomic rules of denitrification in bacterial communities

  NSF · $40k · 2020–2023

Frequent coauthors

• Min Wu

  Tunghai University

  18 shared

• Richard W. Carthew

  14 shared

• Thomas Lecuit

  10 shared

• Boris I. Shraiman

  Princeton University

  10 shared

• Vasyl Alba

  University of Michigan–Ann Arbor

  8 shared

• Muzhi Xu

  Northwestern University

  8 shared

• Pierre‐François Lenne

  Centre National de la Recherche Scientifique

  8 shared

• Michael P. Brenner

  Harvard University

  8 shared

Labs

• Madhav Mani ResearchPI

Awards & honors

• Simons Postdoctoral Fellowship (2010 - 2013)
• Derek Bok Undergraduate Teaching Award, Harvard University (…
• Robert L. Wallace Prize Fellowship, Harvard University (2007…