DBD2BS

DBD2BS predicts DNA-binding specificities by inferring position weight matrices (PWMs) from DNA-binding domains (DBDs) using structural information to identify potential DNA binding sites.

Key Features:

• Prediction from unbound structures: Predicts PWMs from unbound DNA-binding domain (DBD) structures, enabling inference of binding specificity without bound protein–DNA complexes.
• Atom-level knowledge-based potential: Uses an atom-level knowledge-based potential function to estimate changes in binding affinity caused by base substitutions within DNA sequences.
• Extensive template database: Employs a database of 1,066 DBD-DNA complex templates, including 1,813 protein chains, to synthesize bound configurations from unbound queries.
• PWM summarization: Summarizes estimated affinity changes as position weight matrices (PWMs) representing sequence preferences.
• Experimental consistency: Utilizes a kernel that has been validated to produce PWMs consistent with experimental data.
• No-chromatin-immunoprecipitation requirement: Infers PWMs and putative binding sites even in the absence of chromatin immunoprecipitation data.

Scientific Applications:

• Regulatory site identification: Identifies putative binding sites of DNA-binding proteins to support studies of cellular control mechanisms that initiate, enhance, or repress biological processes.
• Genome-wide specificity inference: Enables genome-wide prediction of binding specificities for DBDs when experimental binding data are limited.
• Structure-based binding prediction: Supports prediction of DNA-binding sites from both bound and unbound protein structures.

Methodology:

Uses an atom-level knowledge-based potential to estimate binding-affinity changes due to base substitutions, summarizes these changes as PWMs, and synthesizes bound configurations for unbound queries using the 1,066-template DBD–DNA database.