.png){kind=icon}
.png){kind=icon}
Return
Genomic and evolutionary factors influencing the prediction accuracy of optimal growth temperature in prokaryotes
.png){kind=icon}
PRE
AI
T
M
S
T
I
DOI：10.1128/msystems.00062-26.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Bacteria and archaea have evolved diverse genomic adaptations to thrive across various temperatures. These adaptations include genome sequence optimizations, such as increased GC content in rRNA and tRNA, shifts in codon and amino acid usage, and the acquisition of functional genes conferring adaptation for specific temperatures. Since the experimental determination of optimal growth temperatures (OGTs) is only possible for cultured species, predicting OGT from genomic information has become increasingly important given the exponential increase in genomic data. Although previous studies developed prediction models integrating multiple features based on genome composition using machine learning, the accuracy was variable depending on the target species, with models performing well for thermophiles but less accurately for psychrophiles. In this study, we curated the OGTs and genomic data of 2,869 bacterial species to develop a novel prediction model incorporating features reflecting genomic adaptation toward lower temperatures. We found that species with rapid OGT shifts from their ancestors, including psychrophiles, showed less accuracy in genome composition-based models. Incorporating the gene presence/absence information associated with the rapid changes in OGT improved the prediction accuracy for psychrophiles. We also observed that OGT in archaea is phylogenetically more conserved than in bacteria, which may lead to the long-term optimization of the genome composition and explain high predictability of OGT in archaea. These findings highlight the importance of integrating long- and short-term evolutionary adaptations for phenotype prediction models.IMPORTANCEPrediction of optimal growth temperature (OGT) from genomic data allows for characterizing uncultivated microbes. This study developed a novel OGT prediction model by curating OGT and genomic data of 2,869 bacterial species, demonstrating that the prediction accuracy was improved by incorporating not only genome composition features reflecting long-term temperature adaptation but also the presence/absence information of genes conferring short-term adaptation, particularly for species undergoing rapid OGT shifts. Our study provides a framework for improving phenotype prediction models by integrating long- and short-term evolutionary factors, which may also apply to other microbial physiological phenotypes.
Keywords:
optimal growth temperature
psychrophiles
evolution
machine learning
Journal
IF:
4.6
Papers:
2.8K
Citations:
1.2W
Organization
Cited Papers