Khan Lab
Research programs
We identified downregulated sphingolipid metabolism as an early marker of generalized T2D through comprehensive metabolomics and an unbiased systems biology approach in a prospective cohort study of T2D progression. Building on these findings, we developed a novel bioinformatics tool, metGWAS 1.0, designed to uncover genetic contributors to disease mechanisms. Using metGWAS 1.0, we mapped downregulated sphingolipid species to GWAS databases and identified CERS2 as a key candidate gene and potential functional driver of T2D, harboring the known risk variant rs267738. This SNP, located in exon 3 of the CERS2 gene, results in a point mutation that causes a 33% loss of enzyme function. The association between rs267738 and T2D risk has been consistently replicated across multiple large-scale GWAS and meta-analyses. According to population genetics data from the Ensembl Genome Browser, the rs267738 risk allele occurs in approximately 7% of the population. Despite its strong association with T2D and high population prevalence, the causative role of this CERS2 loss-of-function variant in T2D remained underexplored. Our recent functional studies under non-metabolic stress conditions are the first to demonstrate that CERS2 loss-of-function and the rs267738 variant functionally drive generalized T2D.
Research Highlights
Reduced circulating sphingolipids and CERS2 activity are linked to T2D risk and impaired insulin secretion
Gestational diabetes mellitus (GDM) increases the risk of developing type 2 diabetes (T2D) after pregnancy, though the mechanisms remain unclear. By integrating clinical, metabolomic, and genetic (GWAS) data, this study links reduced sphingolipid biosynthesis, particularly driven by the rs267738 variant in the CERS2 gene, to future T2D risk in women with prior GDM. Experimental mouse models (Cers2 knockout and rs267738 knock-in) showed glucose intolerance and impaired insulin secretion, with isolated islets confirming reduced β-cell function. Overall, decreased sphingolipids may serve as an early biomarker of GDM-to-T2D progression and reflect impaired CERS2-dependent regulation of glucose homeostasis.
The discovery of novel predictive biomarkers and early-stage pathophysiology for the transition from gestational diabetes to type 2 diabetes
This study aimed to identify early predictive markers and uncover mechanisms driving this transition. Using machine learning, a robust seven-lipid metabolite signature was identified that predicts future T2D with high accuracy (AUC 0.92). Pathway analysis showed increased fatty acid metabolism and decreased sphingolipid metabolism in women who later developed T2D. Functional experiments demonstrated that inhibiting sphingolipid synthesis impairs glucose-stimulated insulin secretion in pancreatic islets without affecting whole-body insulin sensitivity. Overall, reduced sphingolipid metabolism emerges as both a strong predictive biomarker and a potential mechanistic contributor to β-cell dysfunction in the progression from GDM to T2D.
Diminished Sphingolipid Metabolism, a Hallmark of Future Type 2 Diabetes Pathogenesis, Is Linked to Pancreatic β Cell Dysfunction
In a matched case-control study of women with prior GDM, postpartum metabolomics showed that reduced sphingolipid metabolism strongly predicts progression to type 2 diabetes (T2D). Integrative metabolomic and genetic analyses identified CERS2 and CERS4 as key genes in this pathway. Functional studies in mice and cells further demonstrated that impaired sphingolipid metabolism disrupts pancreatic β-cell function. Overall, early postpartum defects in sphingolipid biosynthesis may contribute to β-cell dysfunction and increased T2D risk.
metGWAS 1.0: an R workflow for network-driven over-representation analysis between independent metabolomic and meta-genome-wide association studies
Combining genome-wide association studies (GWAS) with metabolomics can identify disease-linked genes and pathways, but typically requires matched datasets from the same individuals—often limiting feasibility. To address this, we developed metGWAS 1.0, a bioinformatics tool that integrates independent GWAS databases with standalone metabolomics data using a network-based systems biology approach. Validation with existing datasets showed that metGWAS successfully recapitulates known associations and uncovers novel gene loci and SNP–metabolite links, enabling disease-relevant insights without the need for paired genomic data.
Other focuses
Disease Heterogeneity
Using a discovery-based unsupervised clustering approach, we analyzed metabolomic, clinical, and biochemical data to uncover early postpartum metabolic heterogeneity in women who developed type 2 diabetes after gestational diabetes.
Deciphering Cellular Cross-Talk in Islets
Integration of AI and traditional medicine in drug discovery
Integrating artificial intelligence (AI) into drug discovery has the potential to transform the development of new therapies, particularly for chronic diseases where effective and safe treatments remain limited. While modern medicines are essential, they often have adverse effects and require continual improvement. Traditional plant-derived medicines offer a valuable but underutilized resource, hindered by outdated exploration methods. This review outlines how AI can be applied across drug discovery stages and proposes a framework for an AI-assisted platform to systematically identify and develop novel therapeutic compounds from plant-based sources.
Omics technologies promised improved biomarker discovery for precision medicine. The foremost problem of discovered biomarkers is irreproducibility between patient cohorts. From a data analytics perspective, the main reason for these failures is bias in statistical approaches and overfitting resulting from batch effects and confounding factors. The keys to reproducible biomarker discovery are: proper study design, unbiased data preprocessing and quality control analyses, and a knowledgeable application of statistics and machine learning algorithms. In this review, we discuss study design and analysis considerations and suggest standards from an expert point-of-view to promote unbiased decision-making in biomarker discovery in precision medicine.
Role of macrophagic MTOR signaling in atherosclerosis
Diet is a key driver of cardiovascular disease (CVD), with growing evidence that excessive dietary protein may contribute to risk. This work demonstrates that high protein intake promotes atherosclerosis through activation of macrophage MTORC1 signaling and suppression of protective autophagy. Translational studies in humans confirm similar effects in monocytes. Mechanistically, leucine emerges as the critical amino acid, triggering pathogenic MTOR activation only above a threshold level. Mouse models further validate that elevated dietary leucine drives atherosclerosis. These findings highlight a novel pathogenic role for leucine in CVD and suggest targeting macrophage leucine–MTOR signaling as a potential therapeutic strategy.
The known mode of action of isoniazid (INH) is to inhibit bacterial cell wall synthesis following activation by the bacterial catalase–peroxidase enzyme KatG in Mycobacterium tuberculosis (Mtb). This simplistic model fails to explain (a) how isoniazid penetrates waxy granulomas with its very low lipophilicity, (b) how isoniazid kills latent Mtb lacking a typical cell wall, and (c) why isoniazid treatment time is remarkably long in contrast to most other antibiotics. To address these questions, a novel comprehensive mode of action of isoniazid has been proposed here based on our studies. Briefly, isoniazid eradicates latent tuberculosis (TB) by prompting slow differentiation of pro-inflammatory monocytes and providing protection against reactive species-induced “self-necrosis” of phagocytes. In the case of active TB, different immune cells form INH-NAD+ adducts to inhibit Mtb's cell wall biosynthesis. This additionally suggests that the antibacterial properties of INH do not rely on KatG of Mtb. As such, isoniazid-resistant TB needs to be re-evaluated.