PRPrajit Rajkumar

Prajit Rajkumar

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Prajit Rajkumar has a total of 184 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Bridging complexity and accessibility in metabolomics with MetaboApps

    by Helena Mannochio-Russo, Wilhan D. Gonçalves Nunes, Haoqi Nina Zhao, Kine Eide Kvitne, Shipei Xing, Harsha Gouda, Julius Agongo, Ipsita Mohanty, Vincent Charron-Lamoureux, Prajit Rajkumar, et al.


  • Open Access

    Alkamines reveal a hidden layer of steroid and drug metabolism

    by Julius Agongo, Subshaskar R Panga, Shipei Xing, Vincent Charron-Lamoureux, Harsha Gouda, Yasin El Abiead, Megan R. Nelson, Abubaker Patan, Marvic Carrillo Terrazas, Kine Eide Kvitne, et al.

    Biomedical research overlooks most genes in favor of a well-studied minority, yet whether analogous blind spots exist in metabolomics remains unknown. We show that reductive amination, forming secondary amines from aldehydes or ketones and amines, generates a previously hidden class of metabolites we term alkamines. Multiplexed synthesis of 8,475 alkamines combined with MS/MS searches across 1.7 billion spectra identified 1,626 candidates across multiple species and organs. Of these, 56 were confirmed in biological samples, including 27 steroid- and 12 drug-derived alkamines matching prescription patterns. Notably, 77% of synthesized alkamines are absent from PubChem. This combinatorial logic likely explains why alkamines have evaded detection and suggests drug metabolism frameworks substantially underestimate drug-derived metabolite diversity. Reductive amination is an overlooked route modifying steroids, bile acids, and xenobiotics.


  • Open Access

    Retrieval-Augmented Language Models Enable Scalable Chemical Source Classification in Metabolomics Workflows

    by Prajit Rajkumar, Runbang Tang, Harshada Sapre, Jasmine Zemlin, Victoria Deleray, Jeong In Seo, Siddharth Mohan, Shipei Xing, Harsha Gouda, Yasin El Abiead, et al.

    There is a growing need for scalable chemical classification to support the interpretation of exposomics and metabolomics data. While structural categorization has been largely automated, functional and exposure-based labeling of chemicals remains a manual and time-consuming process. Here, we present chemsource, a flexible framework that integrates large language models (LLMs) with retrieval-augmented generation (RAG) to automate chemical classification. chemsource retrieves descriptive text from Wikipedia or PubMed abstracts based on chemical names and prompts LLMs to assign user-defined categories based on the retrieved content. We demonstrate classification into five exposure categories: endogenous metabolites, food molecules, drugs, personal care products, industrial chemicals, and combinations of these possibilities. Benchmarking against manually curated labels for 4,953 compounds showed 75% overall agreement, with category-level recall exceeding 75% across all classes. Expert review indicated that most discrepancies could be attributed to prompt ambiguity and incomplete manual labels rather than model failure. To demonstrate the utility of chemsource in metabolomics workflow, we applied it to eight public untargeted metabolomics data sets, revealing distinct exposure patterns across human biospecimens, mouse tissues, environmental dust, and consumer product extracts. chemsource is customizable via prompt editing, enabling diverse classification tasks without requiring coding expertise. The tool is freely available as a Python package (https://pypi.org/project/chemsource/). Text retrieval is free; classification requires user-supplied LLM API access.


  • Open Access

    A resource to empirically establish drug exposure records directly from untargeted metabolomics data

    by Haoqi Nina Zhao, Kine Eide Kvitne, Corinna Brungs, Siddharth Mohan, Vincent Charron-Lamoureux, Wout Bittremieux, Runbang Tang, Robin Schmid, Santosh Lamichhane, Shipei Xing, et al.

    Despite extensive efforts, extracting medication exposure information from clinical records remains challenging. To complement this approach, here we show the Global Natural Product Social Molecular Networking (GNPS) Drug Library, a tandem mass spectrometry (MS/MS) based resource designed for drug screening with untargeted metabolomics. This resource integrates MS/MS references of drugs and their metabolites/analogs with standardized vocabularies on their exposure sources, pharmacologic classes, therapeutic indications, and mechanisms of action. It enables direct analysis of drug exposure and metabolism from untargeted metabolomics data, supporting flexible summarization at multiple ontology levels to align with different research goals. We demonstrate its application by stratifying participants in a human immunodeficiency virus (HIV) cohort based on detected drug exposures. We uncover drug-associated alterations in microbiota-derived N -acyl lipids that are not captured when stratifying by self-reported medication use. Overall, GNPS Drug Library provides a scalable resource for empirical drug screening in clinical, nutritional, environmental, and other research disciplines, facilitating insights into the ecological and health consequences of drug exposures. While not intended for immediate clinical decision-making, it supports data-driven exploration of drug exposures where traditional records are limited or unreliable.


  • Open Access

    Charting the undiscovered metabolome with synthetic multiplexing implicates ibuprofen-carnitine in myotoxicity

    by Abubaker Patan, Shipei Xing, Vincent Charron-Lamoureux, Zhewen Hu, Victoria Deleray, Julius Agongo, Jasmine Zemlin, Harsha Gouda, Prajit Rajkumar, Yasin El Abiead, et al.

    Most molecular features detected in untargeted metabolomics remain uncharacterized due to the limited scope of existing spectral reference libraries. We synthesized >100,000 biologically inspired compounds using multiplexed reactions, of which 91% were absent from existing structural databases, and searched the resulting MS/MS library across >1.7 billion public spectra, increasing annotation rates by 17.4%. This approach revealed previously undescribed exposure-derived metabolites, including ibuprofen–carnitine. Because ibuprofen has been linked to rhabdomyolysis, reduced mitochondrial function, and impaired muscle recovery in carnitine-limited contexts, we investigated the functional relevance of this conjugate. Ibuprofen–carnitine reduced carnitine transport via the OCTN2 transporter, and in a postpartum mouse muscle injury model, ibuprofen delayed muscle repair that could be rescued by carnitine supplementation, with urinary ibuprofen–carnitine:carnitine ratios tracking this effect. These findings support a hypothesis whereby NSAID–carnitine conjugates compete for carnitine transport, impairing energy metabolism and muscle recovery in susceptible individuals. Synthetic multiplexing thus provides a scalable route to annotate the dark metabolome and generate experimentally testable biological hypotheses.


  • Open Access

    Bridging Complexity and Accessibility in Metabolomics with MetaboApps

    by Helena Mannochio-Russo, Wilhan D Gonçalves Nunes, Haoqi Nina Zhao, Kine Eide Kvitne, Shipei Xing, Harsha Gouda, Julius Agongo, Ipsita Mohanty, Vincent Charron-Lamoureux, Prajit Rajkumar, et al.

    Untargeted metabolomics is a powerful approach for exploring the chemical diversity and dynamics of biological systems. However, the types of questions that can be addressed depend not only on experimental design but also on the data processing and analysis workflows employed, many of which require advanced computational expertise. GNPS1, now transitioning to its second major implementation (GNPS2), has evolved into an expandable platform that supports the integration of modular web applications designed to simplify and enhance downstream analysis. These apps, named MetaboApps, facilitate the post-processing of outputs of several GNPS workflows and help make repository-scale metabolomics knowledge and other areas of metabolomics more accessible to a broader community.


  • A guide to reverse metabolomics—a framework for big data discovery strategy

    by Vincent Charron-Lamoureux, Helena Mannochio-Russo, Santosh Lamichhane, Shipei Xing, Abubaker Patan, Paulo Wender Portal Gomes, Prajit Rajkumar, Victoria Deleray, Andrés Mauricio Caraballo-Rodríguez, Kee Voon Chua, et al.

    Untargeted metabolomics is evolving into a field of big data science. There is a growing interest within the metabolomics community in mining tandem mass spectrometry (MS/MS)-based data from public repositories. In traditional untargeted metabolomics, samples to address a predefined question are collected and liquid chromatography with MS/MS data are generated. We then identify metabolites associated with a phenotype (for example, disease versus healthy) and elucidate or validate their structural details (for example, molecular formula, structural classification, substructure or complete structural annotation or identification). In reverse metabolomics, we start with MS/MS spectra for known or unknown molecules. These spectra are used as search terms to search public data repositories to discover phenotype-relevant information such as organ/biofluid distribution, disease condition, intervention status (for example, pre- and postintervention), organisms (for example, mammals versus others), geography and any other biologically relevant associations. Here we guide the reader through a four-part process: (1) obtaining the MS/MS spectra of interest (Universal Spectrum Identifier) and (2) Mass Spectrometry Search Tool searches to find the files associated with the MS/MS that are in available databases, (3) using the Reanalysis Data User Interface framework to link the files with their metadata and (4) validating the observations. Parts 1–3 could take from hours to days depending on the method used for collecting MS/MS spectra. For example, we use MS/MS spectra from three small molecules: phenylalanine-cholic acid (a microbially conjugated bile acid), phenylalanine-C4:0 and histidine-C4:0 (two N-acyl amides). We leverage the Global Natural Products Social Molecular Networking-based framework to explore the microbial producers of these molecules and their associations with health conditions and organ distributions in humans and rodents.


  • Open Access

    To Impute or Not To Impute in Untargeted Metabolomics─That is the Compositional Question

    by Dennis D. Krutkin, Sydney Thomas, Simone Zuffa, Prajit Rajkumar, Rob Knight, Pieter C. Dorrestein, Scott T. Kelley

    Untargeted metabolomics often produce large datasets with missing values. These missing values are derived from biological or technical factors and can undermine statistical analyses and lead to biased biological interpretations. Imputation methods, such as k-Nearest Neighbors (kNN) and Random Forest (RF) regression, are commonly used, but their effects vary depending on the type of missing data, e.g., Missing Completely At Random (MCAR) and Missing Not At Random (MNAR). Here, we determined the impacts of degree and type of missing data on the accuracy of kNN and RF imputation using two datasets: a targeted metabolomic dataset with spiked-in standards and an untargeted metabolomic dataset. We also assessed the effect of compositional data approaches (CoDA), such as the centered log-ratio (CLR) transform, on data interpretation since these methods are increasingly being used in metabolomics. Overall, we found that kNN and RF performed more accurately when the proportion of missing data across samples for a metabolic feature was low. However, these imputations could not handle MNAR data and generated wildly inflated or imputed values where none should exist. Furthermore, we show that the proportion of missing values had a strong impact on the accuracy of imputation, which affected the interpretation of the results. Our results suggest imputation should be used with extreme caution even with modest levels of missing data and especially when the type of missingness is unknown.


  • Open Access

    The microbiome diversifies N -acyl lipid pools - including short-chain fatty acid-derived compounds

    by Helena Mannochio-Russo, Vincent Charron-Lamoureux, Martijn van Faassen, Santosh Lamichhane, Wilhan D. Gonçalves Nunes, Victoria Deleray, Abubaker Patan, Kyle Vittali, Prajit Rajkumar, Yasin El Abiead, et al.

    N -acyl lipids are important mediators of several biological processes including immune function and stress response. To enhance the detection of N -acyl lipids with untargeted mass spectrometry-based metabolomics, we created a reference spectral library retrieving N -acyl lipid patterns from 2,700 public datasets, identifying 851 N -acyl lipids that were detected 356,542 times. 777 are not documented in lipid structural databases, with 18% of these derived from short-chain fatty acids and found in the digestive tract and other organs. Their levels varied with diet, microbial colonization, and in people living with diabetes. We used the library to link microbial N -acyl lipids, including histamine and polyamine conjugates, to HIV status and cognitive impairment. This resource will enhance the annotation of these compounds in future studies to further the understanding of their roles in health and disease and highlight the value of large-scale untargeted metabolomics data for metabolite discovery.


  • Open Access

    To impute or not to impute in untargeted metabolomics - that is the compositional question

    by Dennis Dimitri Krutkin, Sydney Thomas, Simone Zuffa, Prajit Rajkumar, Rob Knight, Pieter C. Dorrestein, Scott T. Kelley

    Untargeted metabolomics often produce large datasets with missing values, arising from biological or technical factors, which can undermine statistical analyses and lead to biased biological interpretations. Imputation methods, such as k-Nearest Neighbors (kNN) and Random Forest (RF) regression are commonly used but their effects vary depending on the type of missing data e.g. Missing Completely At Random (MCAR) and Missing Not At Random (MNAR). Here, we determined the impacts of degree and type of missing data on the accuracy of kNN and RF imputation using two datasets: a targeted metabolomic dataset with spiked-in standards and an untargeted metabolomic dataset. We also assessed the effect of compositional data approaches (CoDA), such as the centered log-ratio (CLR) transform, on data interpretation, since these methods are increasingly being used in metabolomics. Overall, we found that kNN and RF performed more accurately when the proportion of missing data across samples for a metabolic feature was low. However, these imputations could not handle MNAR data and generated wildly inflated values or imputed values where none should exist. Furthermore, we show that the proportion of missing values had a strong impact on the accuracy of imputation which affected the interpretation of the results. Our results suggest extreme caution should be used with imputation even with modestly levels of missing data or when the type of missingness is unknown.