LINflow

LINflow computes genomic similarity matrices for prokaryotic genomes by combining alignment-free and alignment-based approaches to support genome-based classification and identification of strains.

Key Features:

• Alignment-free candidate selection: Uses the alignment-free tool sourmash to identify the genome in a dataset most similar to a given query genome to reduce computational demands.
• Alignment-based ANI calculation: Uses pyani to compute Average Nucleotide Identity (ANI) values between the query genome and the candidate identified by sourmash.
• Hybrid approach: Integrates alignment-free and alignment-based methods to optimize the trade-off between computational efficiency and ANI precision.
• Incremental dataset integration: Applies the same identification and ANI computation process iteratively when adding new genomes to an existing dataset.
• LIN storage and inference: Stores computed ANI values as Life Identification Numbers (LINs) and uses them to infer other pairwise ANI values within the dataset.
• Performance benchmarking: Demonstrated up to 150-fold speed improvement over pyani on four datasets totaling 484 genomes while maintaining high correlation with pyani-computed ANI values, with occasional minimal discrepancies.

Scientific Applications:

• Genome-based classification and strain identification: Generates ANI matrices to support classification and identification of prokaryotic strains.
• Continuous dataset updating: Rapidly integrates newly sequenced genomes into existing similarity matrices for ongoing genomic surveillance or comparative studies.
• Large-scale ANI estimation and inference: Enables efficient estimation of ANI and inference of pairwise similarities across large prokaryotic genome collections.

Methodology:

LINflow first uses sourmash (alignment-free) to find the most similar genome candidate for a query, then uses pyani (alignment-based) to compute ANI between the query and that candidate; computed ANI values are stored as Life Identification Numbers (LINs) and the same process is applied iteratively when adding new genomes, enabling inference of other pairwise ANI values.