DynaSig-ML

DynaSig-ML predicts how sequence variants affect 3D structural dynamics of biomolecules and uses those dynamics as input features for machine learning models to predict experimental outcomes.

Key Features:

• Integration with Machine Learning: Accepts user-selected machine learning models to interpret datasets of experimental measures derived from numerous sequence variants.
• Elastic Network Contact Model (ENCoM): Uses ENCoM, a sequence-sensitive coarse-grained normal mode analysis model, to predict 3D structural dynamics for each variant.
• Dynamical Signatures: Generates Dynamical Signatures that capture positional fluctuations at every position within a biomolecule for use as machine-learning features.
• Prediction of Experimental Outcomes: Trained models predict experimental outcomes for new or theoretical sequence variants.
• Computational Efficiency and Parallelization: Compute-intensive steps can be parallelized to handle large biomolecules and extensive sequence variant datasets.

Scientific Applications:

• Dynamics–Function Relationship Analysis: Exploration of how structural dynamics influence biological function across biomolecules.
• Variant Effect Prediction: Prediction of the effects of sequence variants on experimental measures and molecular function.
• High-Throughput Assay Interpretation: Interpretation of high-throughput enzymatic assay data, for example predicting maturation efficiency of human microRNA miR-125a variants.

Methodology:

Predicts 3D structural dynamics using ENCoM (sequence-sensitive coarse-grained normal mode analysis). Generates Dynamical Signatures representing positional fluctuations. Uses these signatures as input features for user-selected machine learning models which are trained to predict experimental outcomes for new or theoretical variants. Compute-intensive steps can be parallelized.