SIDR

SIDR uses decision-tree machine learning to identify and remove contaminant sequences from Next Generation Sequencing (NGS) de novo genome assemblies to improve assembly quality, particularly for eukaryotic genomes.

Key Features:

• Machine Learning Approach: Decision tree algorithm for sequence classification that handles diverse input features without requiring prior identification of significant descriptors.
• Feature Flexibility: Can incorporate any measured feature as input, enabling classification based on diverse sequence- and assembly-derived characteristics.
• Contaminant Identification: Distinguishes target-organism sequences from contaminants, addressing cases in eukaryotic assemblies where nucleotide similarity can cause target sequence loss.
• Comparative Superiority: Decision trees used by SIDR have demonstrated superior performance compared to existing protocols in classifying sequences within eukaryotic de novo assemblies.

Scientific Applications:

• Decontamination of Eukaryotic Assemblies: Purges contaminant sequences from DNA extracts and NGS datasets containing non-target organism sequences.
• Enhanced Genome Assembly Quality: Improves the quality and reliability of de novo genome assemblies by filtering contaminant sequences.

Methodology:

SIDR trains decision trees on sequence data using features derived from classifiers such as BLAST to classify sequences as target or contaminant.