SATIVA

SATIVA detects taxonomically mislabeled molecular sequences and proposes corrected taxonomic classifications by assessing phylogenetic placement using the Evolutionary Placement Algorithm (EPA) and statistical models of evolution to improve accuracy in reference databases and downstream metagenomic analyses.

Key Features:

• Phylogeny-Aware Detection: Uses the Evolutionary Placement Algorithm (EPA) to evaluate congruence between sequence annotations and their placement in a reference phylogeny.
• Statistical Modeling and Scoring: Employs statistical models of evolution to score sequences based on phylogenetic signal support for their taxonomic labels.
• Automated Correction Proposals: Generates proposed corrected taxonomic classifications for sequences identified as mislabeled.
• Detection Accuracy: Reported detection sensitivity of 96.9% and precision of 91.7% on simulated data.
• Correction Accuracy: Reported correction sensitivity of 94.9% and precision of 89.9% on simulated data.

Scientific Applications:

• Quality Improvement of Reference Databases: Identifies and corrects mislabels in microbial 16S reference databases such as Greengenes, LTP, RDP, and SILVA to enhance database reliability.
• Metagenomic Studies: Reduces bias from taxonomic inaccuracies to improve the validity of metagenomic analyses.
• Taxonomy Evaluation: Facilitates evaluation of alternative taxonomies, demonstrated for Cyanobacteria.

Methodology:

Places sequences with the Evolutionary Placement Algorithm (EPA), scores taxonomic congruence using statistical models of evolution, detects sequences with low phylogenetic support for their labels, and proposes automated corrected classifications.