odseq

odseq identifies outlier sequences in multiple sequence alignments (MSAs) or unaligned sequence sets by computing average pairwise distances to improve alignment quality and downstream analyses.

Key Features:

• Outlier Detection Algorithm: Computes the average distance of each sequence to all others and flags sequences with anomalous distances as outliers.
• Distance Metrics and Methods: Uses predefined distance metrics and determines anomalous values with statistical methods such as the interquartile range or bootstrapping.
• Scalability and Efficiency: Applies the mBed algorithm from Clustal Omega to reduce computational complexity from O(N²) to O(N log N), enabling processing of large sequence sets on a single core.
• Versatility: Operates on MSAs and on sets of unaligned sequences, with reduced accuracy on unaligned data relative to aligned datasets.
• Performance: Validated on artificial test cases using sequences from Pfam families mixed with sequences from other families, demonstrating accurate outlier identification and rapid processing of medium-sized alignments (thousands of sequences) in seconds.

Scientific Applications:

• Alignment curation: Detects and enables removal of outlier sequences to improve multiple sequence alignment integrity.
• Phylogenetics: Reduces inclusion of spurious sequences prior to tree inference to improve phylogenetic analyses.
• Structural biology: Improves sequence sets used for homology modeling and conservation analyses in structural studies.
• Evolutionary studies: Enhances detection of homologous relationships and family-specific analyses by excluding divergent or non-homologous sequences.

Methodology:

Accepts an MSA, distance matrix, or set of unaligned sequences; computes the average pairwise distance for each sequence; identifies outliers by statistical analysis of distance distributions using the interquartile range or bootstrapping; and can use the mBed algorithm from Clustal Omega to approximate distances and reduce complexity to O(N log N).