adabmDCA 2.0-A Flexible but Easy-to-Use Package for Direct Coupling Analysis

Area of Science:

• Computational Biology
• Bioinformatics
• Machine Learning

Background:

• Direct Coupling Analysis (DCA) is a powerful method for inferring biological sequence information.
• Existing DCA implementations can be complex and lack flexibility for diverse applications.
• There is a need for user-friendly and versatile DCA tools applicable to various biological sequences and computational architectures.

Purpose of the Study:

• To introduce adabmDCA 2.0, a novel implementation of direct coupling analysis (DCA).
• To provide a flexible and easy-to-use tool for various downstream tasks in sequence analysis.
• To support multiple programming languages and computational architectures.

Main Methods:

• Implementation of DCA using Boltzmann machine learning.
• Development of a common front-end interface for C++, Julia, and Python.
• Inclusion of various learning protocols for dense and sparse generative DCA models.
• Support for single-core, multicore CPU, and GPU architectures.

Main Results:

• adabmDCA 2.0 offers a flexible and user-friendly implementation of DCA.
• The package supports multiple programming languages and diverse computational hardware.
• It enables direct application to residue-residue contact prediction, mutational-effect prediction, and sequence library scoring.
• The tool facilitates the generation of artificial sequences for protein and RNA design.

Conclusions:

• adabmDCA 2.0 provides a versatile and accessible platform for advanced sequence analysis using DCA.
• The package empowers researchers in protein and RNA sequence data analysis and design.
• This implementation simplifies complex DCA tasks, enhancing discoverability in biological sequence research.