BiDaS

BiDaS generates Monte Carlo–simulated DNA/RNA and amino acid sequences together with numerical feature datasets that reproduce original data distributions and preserve between-feature associations for bioinformatics analyses.

Key Features:

• Monte Carlo Simulation: BiDaS employs Monte Carlo sampling to generate simulated sequences and numerical features that match the distributions of input data.
• Sequence Generation: Simulates DNA/RNA and amino acid sequences according to the sequence and feature distributions of user-provided datasets for studies of gene, transcript, peptide, and protein groups/families.
• Numerical Feature Simulation: Creates numerical features with identical distributional properties to the original data while maintaining precise 2D and 3D correlations between features.
• Multidimensional Feature Projection: Projects sequences into multidimensional feature spaces using 38 DNA/RNA features from the B-DNA-VIDEO database, 122 DNA/RNA features from the DiProDB database, and pseudo-amino acid composition derived from input sequences.

Scientific Applications:

• Family analysis: Enables expansion of limited datasets to conduct in-depth studies of gene, transcript, peptide, and protein families.
• Machine learning: Produces extensive realistic training datasets to improve and benchmark machine learning classifiers.
• Hypothesis testing and model validation: Simulates sequences with defined properties to explore hypothetical scenarios and validate computational models.

Methodology:

Uses user-provided small datasets as templates and Monte Carlo sampling to generate large simulated datasets, projects sequences into feature spaces using B-DNA-VIDEO and DiProDB features and pseudo-amino acid composition, and preserves between-feature associations including 2D and 3D correlations.