Tally

Tally identifies structural tandem repeats (TRs) in protein sequences using machine learning to distinguish sequences with genuine 3D repeats from those without.

Key Features:

• Machine Learning Approach: Tally employs machine learning algorithms to detect structural tandem repeats in protein sequences.
• Detection of Imperfect and Evolutionarily Obscured TRs: The method targets imperfect and evolutionarily modified repetitive patterns that are difficult to detect by sequence analysis alone.
• Benchmark Performance: Tally achieved sensitivity of 81%, specificity of 74%, and an Area Under the Receiver Operating Characteristic Curve (AUC) of 86%.
• Structural Benchmarking: Evaluation uses protein 3D structures composed of repetitive 3D blocks as a benchmark to assess detection of structural TRs.
• Improved Separation from Traditional Scoring: The approach provides superior separation between sequences with and without 3D TRs compared to traditional scoring methods.

Scientific Applications:

• Protein Structure Analysis: Identification of structurally significant TRs to inform protein architecture and function.
• Functional Annotation: Selection of functionally meaningful TRs from proteomes to support annotation efforts.
• Benchmarking Resource: Generation of a dataset for benchmarking detection of structural TRs in proteins.

Methodology:

Analyzes protein sequences and integrates machine learning models trained on known 3D protein structures to distinguish sequences with genuine structural TRs from those without.