Deciphering the microbial complexity of Chinese traditional sourdough through integrated learning

Liu, Yan, Wang, Jinyu, Zhou, Shiwei, Li, Feijiang, Cheng, Honghong, Zhu, Zheqing, Wang, Jiale, Liu, Aowen, Lu, Ting, Yu, Yujuan, Tian, Senmiao, Zhang, Min, Sadiq, Faizan Ahmed ORCID: https://orcid.org/0000-0003-1596-4155 and Zhang, Guohua 2026. Deciphering the microbial complexity of Chinese traditional sourdough through integrated learning. Food Research International 225 , 117992. 10.1016/j.foodres.2025.117992

PDF - Accepted Post-Print Version Available under License Creative Commons Attribution. Download (1MB)

Abstract

Sourdough fermentation relies on a dynamic microbial consortium of fungi and bacteria, shaped by interspecies interactions and environmental conditions. We analyzed the microbiota and physicochemical characteristics of 115 sourdough samples – including both 26 dry samples and 89 wet samples – collected across ten Chinese provinces, and used integrated learning approaches to correlate microbial composition with physicochemical properties, revealing associations that underpin fermentation outcomes and regional variation in community structure. In dry samples, Pediococcus pentosaceus(39.46 %)and Levilactobacillus brevis(11.02 %)dominated, whereas wet samples were dominated by Fructilactobacillus sanfranciscensis (51.27 %). Saccharomyces cerevisiae was recognized as a dominant and core yeast species (detected in over 50 % of all samples and ranking among the top 5 in any groups) in the sourdough microbiota. Saccharomycopsis fibuligera, Lactiplantibacillus plantarum and Leuconostoc mesenteroides were core bacterial species. Ensemble learning and abundance ranking identified rare taxa such as Dermacoccus nishinomiyaensis, Kodamaea ohmeri and Buckleyzyma phyllomatis that, despite their low abundance in the sourdough microbiota, showed significant correlations with physicochemical indices, although their functional roles remain uncharacterised. Microbial community structure correlated strongly with physicochemical conditions, including moisture content, pH and total acidity, underscoring the importance of considering these parameters when designing stable synthetic consortia for sourdough fermentation. This study provides foundational insights into the structure and function of sourdough microbiomes across diverse regions and sample types. By integrating microbiota data with physicochemical characteristics, we demonstrate the potential of machine learning to uncover key microbial-environment relationships. These findings support the rational design of stable, tailored starter cultures for improved sourdough fermentation.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Schools > Dentistry
Additional Information:	RRS applied 19/12/2025 AB
Publisher:	Elsevier
ISSN:	0963-9969
Date of First Compliant Deposit:	19 December 2025
Date of Acceptance:	3 December 2025
Last Modified:	19 Dec 2025 14:15
URI:	https://orca.cardiff.ac.uk/id/eprint/183386

Actions (repository staff only)

Edit Item