Building a tRNA thermometer to estimate microbial adaptation to temperature

Area of Science:

• Microbiology
• Genomics
• Computational Biology

Background:

• Ambient temperature influences biochemical reactions, necessitating protein adaptations in extremophiles.
• Predicting optimal growth temperature (OGT) for all microbes is challenging, but genomic differences offer clues.
• Organisms adapted to different temperatures exhibit distinct DNA, RNA, and protein compositions.

Purpose of the Study:

• To develop a novel 'tRNA thermometer' model for predicting microbial optimal growth temperature (OGT) using only tRNA sequences.
• To leverage Convolutional Neural Networks (CNNs) for accurate OGT prediction from genomic data.
• To create a simplified OGT prediction method with minimal data requirements.

Main Methods:

• Trained two CNN models using tRNA sequences from 100 archaea and 683 bacteria.
• Model 1: Paired individual tRNA sequences to predict thermophilic origin (accuracy 0.538-0.992).
• Model 2: Utilized complete species tRNA sets for OGT prediction (max r² of 0.86).

Main Results:

• The tRNA thermometer model achieved high accuracy in predicting thermophilic origins.
• The species-level model demonstrated strong performance in predicting OGT, comparable to existing methods.
• The model requires significantly less input data and reduces preprocessing steps compared to prior approaches.

Conclusions:

• The tRNA thermometer offers an efficient and accurate method for predicting microbial OGT from genome sequences.
• This approach simplifies OGT prediction, removing the need for laborious feature extraction.
• The model broadens the possibilities for downstream analyses in microbial adaptation and evolution studies.