Focused natural product elucidation by prioritizing high-throughput metabolomic studies with machine learning

Tobias, N.; Parra-Rojas, C.; Shi, Y.-N.; Shi, Y.-M.; Simonyi, S.; Thanwisai, A.; Vitta, A.; Chantratita, N.; Hernandez-Vargas, E. A.; Bode, H. B.

doi:10.1101/535781

bioRxiv · 10.1101/535781

Focused natural product elucidation by prioritizing high-throughput metabolomic studies with machine learning

Tobias, N.·Parra-Rojas, C.·Shi, Y.-N.·Shi, Y.-M.·Simonyi, S.·Thanwisai, A.·Vitta, A.·Chantratita, N.·Hernandez-Vargas, E. A.·Bode, H. B.

View original source ↗Export metadata ↓

Abstract

Bacteria of the genera Photorhabdus and Xenorhabdus produce a plethora of natural products to support their similar symbiotic lifecycles. For many of these compounds, the specific bioactivities are unknown. One common challenge in natural product research when trying to prioritize research efforts is the rediscovery of identical (or highly similar) compounds from different strains. Linking genome sequence to metabolite production can help in overcoming this problem. However, sequences are typically not available for entire collections of organisms. Here we perform a comprehensive metabolic screening using HPLC-MS data associated with a 114-strain collection (58 Photorhabdus and 56 Xenorhabdus) from across Thailand and explore the metabolic variation among the strains, matched with several abiotic factors. We utilize machine learning in order to rank the importance of individual metabolites in determining all given metadata. With this approach, we were able to prioritize metabolites in the context of natural product investigations, leading to the identification of previously unknown compounds. The top three highest-ranking features were associated with Xenorhabdus and attributed to the same chemical entity, cyclo(tetrahydroxybutyrate). This work addresses the need for prioritization in high-throughput metabolomic studies and demonstrates the viability of such an approach in future research.

Explore related subjects

microbiology Preprint

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Tobias, N., Parra-Rojas, C., Shi, Y.-N., Shi, Y.-M., Simonyi, S., Thanwisai, A., Vitta, A., Chantratita, N., Hernandez-Vargas, E. A., Bode, H. B.. 2019-01-31. Focused natural product elucidation by prioritizing high-throughput metabolomic studies with machine learning. https://doi.org/10.1101/535781

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Cell-scale confinement reorders bacterial speed rankings

Which bacterium swims fastest depends on where the race is run. Speeds are usually measured in open liquid, yet many bacteria live in pores, mucus and host passages no wider than a cell. Using a microfluidic device with eight channel widths (0.80-3.00 m) in one field of view, we tracked unlabeled cells of eight species spanning five motility systems. Chamber height, a polymer additive and channel width each reordered the ranking. Vibrio alginolyticus, the fastest in open liquid, slowed to ~4 m/s in 1.25-m channels, whereas Campylobacter jejuni and Bacillus subtilis kept 33-46 m/s. Open liquid speed did not predict robustness to confinement. In two unrelated lineages, host symbionts that pass through narrow passages during colonization, Caballeronia insecticola and Vibrio fischeri, outran their non-symbiotic relatives in the channels, and crossed them with flagella wrapped around the cell body at nearly the speed of unwrapped cells. Where space is tight, the ability to keep moving, not top speed, sets the ranking.

microbiology↗

bioRxiv2026

The biochemical composition of coronavirus Nucleocapsid proteins drives the bicistronic properties of their sub-genomic mRNAs

The nucleocapsid (N) gene of coronaviruses (CoVs) encodes a structural RNA-binding protein required for viral assembly. In SARS- and MERS-CoV viruses, reading frame 2 of N contains an overlapping open reading frame (ORF) encoding an innate immune antagonist accessory protein. However, the extent of this dual coding capacity across the CoV family, and the genetic features driving overlapping ORF formation, remain unexamined. Here, we developed an in-silico tool (ORFan) to screen structural gene sub-genomic mRNA sequences from 152 species of coronavirus for the presence of overlapping accessory gene candidates. We found that frame 2 of the N gene in Alpha-, Beta- and Delta-CoVs frequently contains a potential accessory ORF (AcORF), conserved at the subgenus level, unlike the spike, envelope and membrane genes. We also show that CoV nucleotide biases in conjunction with codon biases integral to N protein function drive the relatively high abundances of start codons in frame two and stop codons in frame three, explaining the tendency towards frame 2 AcORFs. We further suggest that viral nucleotide biases and the codon composition required for RNA-binding function facilitate overlapping ORF formation in other virus families

microbiology↗

bioRxiv2026

Evaluating non-oxidative glycolysis using 13C metabolic flux analysis

Non-oxidative glycolysis (NOG) converts glucose to acetyl-CoA without emitting CO2 during Escherichia coli bioproduction. Although 13C-metabolic flux analysis (13C-MFA) has the potential to evaluate NOG flux distribution in vivo, the precision of these estimates depends on the labeling patterns of glucose tracers. Because E. coli has multiple native glycolysis pathways, selecting the optimal glucose tracer is important for precise estimates. This study used simulations to evaluate the performance of 18 commercially available glucose tracers in precisely predicting NOG flux. We artificially generated flux distributions of central metabolic pathways, including three types of NOG pathways, and obtained 13C labeling patterns of amino acids metabolized from each glucose tracer. Using these data, we then calculated 95% confidence intervals (CIs) for NOG pathway fluxes. Comparing the sum of CI widths from three NOG fluxes showed that estimation with [1,3-13C]-glucose yielded the narrowest CIs. Several tracers, including [2,3-13C]glucose, were capable of determining fluxes with high precision, but [1-13C]glucose was not one of them, despite often being used for estimating glycolytic fluxes. We then experimentally evaluated [2,3-13C]- and [1-13C]glucose as tracers in E. coli strains with and without the expression of phosphoketolase, which is required for NOG pathways to function. The results suggested that [1-13C]glucose could not detect NOG fluxes, whereas [2,3-13C]glucose estimated the fluxes of all three NOG pathways with narrow CIs.

microbiology↗