ALICO

ALICO generates randomized versions of input multiple sequence alignments while preserving dependence structure, average percent identities (PID), and per-sequence k-mer composition to produce alignment-constrained null models for statistical validation of sequence analyses.

Key Features:

• Preservation of Crucial Alignment Features: Randomized alignments retain the dependence structure present in the original multiple sequence alignment.
• Percent Identity (PID) Maintenance: The method preserves average pairwise percent identities between input sequences.
• K-mer Composition Resemblance: Per-sequence k-mer composition is maintained to mirror genomic training data characteristics.
• Pairwise Alignment Training Data Requirement: The approach operates using only pairwise alignment training data rather than multiple-alignment training sets.

Scientific Applications:

• Homology-aware Finders: Provides null models for homology-aware tools such as PhyloCon, MEME with conservation prior, and PRIORITY-C.
• Naive Finders: Supplies randomized alignments for naive finders like GibbsMarkov and for datasets such as the MacIsaac orthologous yeast data.
• Combining Results from Multiple Finders: Enables combination of results via p-values derived from ALICO sampling to integrate outputs from multiple sequence analysis tools.

Methodology:

ALICO generates alignment-constrained null models by systematic randomization of sequences while maintaining dependence structure, average PID, and k-mer composition, and produces samples used to compute p-values for statistical validation.