I aim to formalize heterogeneity as a measurable and modelable property of disease. By linking molecular network perturbations to cellular programs, patient stratification, and population-level representations, I seek to build predictive, mechanism-aware frameworks for complex diseases.
Structured heterogeneity emerges from modular perturbations across scales.
I build computational frameworks to model phenotypic heterogeneity across molecular, cellular, patient, and population scales.
Disease heterogeneity is not noise - it encodes structured biological states.
My research asks how modular perturbations in molecular networks propagate upward to shape cellular programs, define patient subgroups, and structure population-level variation. Across neurodegeneration, critical illness, inflammatory disease, and oncology, the unifying objective is to extract reproducible disease architecture from high-dimensional data.
My recent work has focused on identifying molecular states that stratify complex clinical syndromes.
In Alzheimer’s disease, large-scale untargeted metabolomics of 500 postmortem brain samples demonstrated that tau pathology (rather than beta-amyloid) is the dominant driver of metabolic dysfunction (Batra*, Arnold* et al., 2022). Comparative analyses across tauopathies further revealed convergent metabolic signatures across AD and PSP (Batra*, Krumsiek*, Wang* et al., 2024).
In acute respiratory distress syndrome (ARDS), integrated multi-omic profiling distinguished molecular subtypes across COVID-19 and bacterial sepsis etiologies, identifying signatures of vascular injury and mortality-associated signatures (Price*, Benedetti*, et al., 2022; Buyukozkan*, Alvarez-Mulett*, et al., 2022; Batra*, Whalen*, et al., 2022; Batra*, Uni*, et al., 2023).
A randomized dietary intervention study demonstrated that Alzheimer’s-associated metabolic states are modifiable, with coordinated serum–CSF remodeling under a ketogenic regimen (Schweickart*, Batra* et al., 2024).
Clinical syndromes comprise biologically distinct molecular states that can be detected and, in some cases, modulated.
To extract coherent biological structure from heterogeneous datasets, I have developed and benchmarked network-guided algorithms for multi-omics integration.
Representative contributions include:
Robust mixture modeling for longitudinal mixed-type clinical data (Haggenberg*, Budde*, ... Batra†, et al., 2024)
Clinical-phenotype–guided transcriptomic modeling (Batra*†, Stark*, Lauffer*, Jargosch* et al., 2021)
Systematic benchmarking of module detection methods (Batra† et al., 2017)
Connected subnetwork extraction across omics layers (Alcaraz N*, Pauling J*, Batra R*, et al., 2014)
Biological meaning emerges from structured integration, not isolated features.
My early work in pediatric neuroblastoma applied time-resolved imaging and functional genomics to identify therapeutic vulnerabilities in aggressive tumors.
By modeling mitotic checkpoint perturbations, we identified dependency on MAD2L1 in p53–p21–deficient neuroblastoma cells (Batra et al., 2012; Gogolin et al., 2013; Harder et al., 2015).
Causal vulnerabilities are revealed when dynamic perturbations are modeled explicitly.