Int&in: A machine learning-based web server for active split site identification in inteins

Area of Science:

• Biochemistry
• Molecular Biology
• Bioinformatics

Background:

• Inteins are protein elements that catalyze their own excision and ligation of flanking protein sequences (protein splicing).
• Inteins exist as contiguous or split forms; split inteins require two fragments to associate for splicing.
• Artificial splitting of inteins enables novel applications in synthetic biology and protein interaction studies.

Purpose of the Study:

• To develop a predictive model for identifying active and inactive intein split sites.
• To create a user-friendly web server (Int&in) for predicting intein split sites.
• To facilitate the engineering of novel split inteins with tailored properties.

Main Methods:

• Development of a machine learning model using logistic regression.
• Training the model on 126 split sites from gp41-1, Npu DnaE, and CL inteins.
• Validation using 97 split sites from scientific literature.
• Incorporation of protein structural features and sequence conservation data.

Main Results:

• The machine learning model achieved high accuracy in predicting active and inactive split sites (0.79 training, 0.78 testing).
• The Int&in web server provides accurate predictions for intein split site activity.
• The model demonstrates the utility of structural and sequence information for predicting intein behavior.

Conclusions:

• The Int&in web server accurately predicts intein split sites, overcoming heuristic identification methods.
• This tool will accelerate the design and engineering of split inteins for diverse applications.
• The findings support the use of machine learning in protein engineering and synthetic biology.