ARTEMIS: an Information-Theory Based Framework for Analysis of Communication in Biomolecular Systems

Cellular processes rely on interactions of biomolecules, i.e. direct contact and subsequent configurational changes that provide a physical basis for information processing and transmission within the cell as well as for communication between its remote parts and with the environment. Molecular communication persists also at the level of a single biomolecule, whereby it is known as allostery - an internal mechanism to regulate biomolecular function. To investigate communication phenomena in different (bio)molecular systems, a universal information-theoretical approach utilizing mutual information (MI) was developed and implemented as a Python/C++ software package ARTEMIS. This tool gives the possibility to analyze intra- and intermolecular communication networks using all-to-all MI matrix for amino acids in proteins, nucleotides in nucleic acids, or any other molecular subgroups defined in the system. In the present study, configurational ensembles from microsecond all-atom molecular dynamics (MD) simulations are subjected into PARENT software to extract MI values from pairwise distributions of internal degrees of freedom and further processed in ARTEMIS to build the thus derived MD MI-matrices, remove noise, test convergence and analyze communication networks. Alternatively, for proteins and nucleic acids MI matrices can be generated from multiple sequence alignments (MSA) and further processed in the same way using ARTEMIS. Specifically, the tool allows: a) construction and visualization of MI matrices; b) analysis of pairwise communication and corresponding sequence biases (e.g. “specific contacts”); c) reconstruction of molecular communication networks and analysis of their modular organization using different clustering algorithms; d) search of allosterically coupled sites for a given target subset (e.g. active site); e) identification of critical residues for information transmission between communicating sites (e.g. active site / allosteric site). A potentially wide application of ARTEMIS is shown for different molecular systems, including lipid bilayers, peptides, disordered and globular proteins. Moreover, analysis of allosteric communication in proteins is shown to be in good agreement with the available experimental data. The MI-based approach realized in ARTEMIS provides a universal and efficient method for analysis of communication in any (bio)molecular system.