Machine learning reveals proteome-encoded growth determinants underlying metabolic versatility of Rhodopseudomonas palustris on lignin-derived aromatics
mSystems · 23.06.2026 · 11 (6)
DOI: 10.1128/msystems.00383-26 ↗Kurz gefasst
Fachartikel in der Zeitschrift mSystems, erschienen am 23.06.2026. Verfasst von Abraham Osinuga und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Chemical and Biomolecular Engineering, University of Nebraska-Lincoln, Lincoln, Nebraska, USA. Hauptthemen laut MEDLINE: Rhodopseudomonas, Lignin, Machine Learning, Proteome. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 106 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NIGMS NIH HHS, NIH HHS, National Science Foundation.
Abstract (Originalsprache)
Lignin depolymerization generates mixtures of aromatic compounds that are promising carbon sources for microbial bioconversion, yet the constraints governing microbial growth on these substrates remain unclear. Here, we investigated how Rhodopseudomonas palustris CGA009 organizes metabolism to grow on chemically distinct lignin-derived aromatics under aerobic and anaerobic conditions. Quantitative proteomics across 14 substrate-oxygen environments revealed extensive oxygen-dependent proteome remodeling, consistent with shifts between respiratory and photoheterotrophic programs. Despite this reorganization, predictive modeling showed that growth rate is encoded by a relatively small subset of proteins whose abundance tracks physiological performance across environments. Using CorePredX, a machine-learning framework that combines global importance analysis with dependence-aware redundancy filtering, we identified 118 high-confidence proteins whose variation made conditionally non-redundant contributions to growth-rate prediction among 1,857 quantified proteins. These proteins are organized into a hierarchical architecture comprising a cross-condition core, adaptive regulators, substrate-specific specialists, and conditional hubs. The cross-condition core linked translational capacity, sulfur assimilation, redox buffering, and carbon storage cycling, highlighting conserved proteomic features associated with growth across oxygen regimes and substrate chemistries. Notably, an uncharacterized cystathionine beta-synthase domain protein, RPA3416, emerged as a strong predictive component of this core, raising the hypothesis of an adenylate-responsive process associated with aerobic growth on lignin-derived aromatics. Together, these results suggest that the metabolic versatility of R. palustris arises from flexible regulation layered onto partially conserved proteome-encoded growth-associated constraints, providing candidate targets for lignin bioconversion engineering and metabolic model refinement. IMPORTANCE Lignin is an abundant part of plant waste that is notoriously difficult to break down and turn into valuable products. While certain microbes, like Rhodopseudomonas palustris , can consume lignin byproducts, scientists do not fully understand how they do it. In this study, we discovered that we can predict how well this microbe will grow simply by tracking a small, specific group of its core proteins—even when its environment changes. We also identified a new, unstudied protein that is a strong candidate for future research. By pinpointing the specific biological tools microbes use to process lignin, this research provides a clear roadmap for engineering bacteria to efficiently turn plant waste into useful, sustainable materials.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Department of Chemical and Biomolecular Engineering, University of Nebraska-Lincoln, Lincoln, Nebraska, USA.
ORCID-Profile: Abraham Osinuga, Rajib Saha
Förderung
- NIGMS NIH HHS
- NIH HHS
- National Science Foundation
- National Institutes of Health
Zitieren
APA
Osinuga, A., Kathol, M., & Saha, R. (2026). Machine learning reveals proteome-encoded growth determinants underlying metabolic versatility of Rhodopseudomonas palustris on lignin-derived aromatics. mSystems, 11(6). https://doi.org/10.1128/msystems.00383-26
Vancouver
Osinuga A, Kathol M, Saha R. Machine learning reveals proteome-encoded growth determinants underlying metabolic versatility of Rhodopseudomonas palustris on lignin-derived aromatics. mSystems. 2026;11(6). doi:10.1128/msystems.00383-26
BibTeX
@article{Osinuga2026,
author = {Osinuga, Abraham and Kathol, Mark and Saha, Rajib},
title = {Machine learning reveals proteome-encoded growth determinants underlying metabolic versatility of Rhodopseudomonas palustris on lignin-derived aromatics},
journal = {mSystems},
year = {2026},
volume = {11},
number = {6},
doi = {10.1128/msystems.00383-26}
}
EUVIMED