MASSIF

MASSIF enhances the accuracy of linking DNA-binding motifs to transcription factors by incorporating protein DNA-binding domain (DBD) information to validate motif–TF associations for studies of transcriptional regulation.

Key Features:

• DBD-based validation: Uses DNA-binding domain (DBD) information to assess whether a motif and a TF are likely associated via similar DBDs.
• Curated TF–motif collection: Constructs a specialized collection of TFs with known DBDs and their corresponding motifs to enable DBD comparisons.
• Domain score (P-value): Quantifies the likelihood that a linked motif and TF share the same DBD using a domain score represented as a P-value.
• Meta-analysis approach: Combines P-values from existing motif–TF linking tools based on TF-associated sequences with the domain score P-value for a joint statistical assessment.
• Filtering approach: Filters out unlikely motif–TF associations by applying the domain score to refine candidate links.
• Evaluation on empirical data: Demonstrated performance using human ChIP-seq datasets with motifs from HOCOMOCO and de novo identified sequences.
• DBD-agnostic improvement: Reported to improve motif–TF linkage performance across different DBD types.

Scientific Applications:

• ChIP-seq motif annotation: Enhances assignment of motifs to candidate TFs in analyses of TF-associated sequences from ChIP-seq.
• Motif–TF association validation: Provides an orthogonal DBD-based criterion to validate computational motif–TF links.
• Gene regulatory network inference: Supports more accurate identification of TFs underlying motif occurrences for studies of transcriptional regulation and regulatory network reconstruction.
• Motif filtering and prioritization: Refines sets of candidate motif–TF associations by removing biologically unlikely links based on DBD compatibility.

Methodology:

Constructs a curated set of TFs with known DBDs and corresponding motifs, computes a domain score as a P-value measuring the likelihood that a motif and TF share the same DBD, and applies either a meta-analysis that combines existing motif–TF linking P-values with the domain score or a filtering strategy that removes associations based on the domain score; performance was evaluated on human ChIP-seq datasets using HOCOMOCO and de novo motifs.