VALENCIA

VALENCIA classifies human vaginal microbial communities into Community State Types (CSTs) by comparing taxonomic composition to predefined reference centroids using a nearest-centroid algorithm.

Key Features:

• Nearest-centroid classification: Assigns samples to CSTs using a nearest-centroid algorithm based on taxonomic composition.
• Reference centroids: Uses predefined reference centroids derived from 13,160 taxonomic profiles collected from 1,975 women in the United States.
• Robustness across datasets: Validated on three distinct test datasets including reproductive-age African women, adolescent girls, and menopausal women.
• Tolerance of technical variation: Demonstrates consistent performance despite variation in sequencing strategies and bioinformatics pipelines.
• Expanded CST taxonomy: Identifies and characterizes CSTs prevalent among reproductive-age women, extending previously defined CST categories.
• Links to clinical metrics: Enables exploration of relationships between CSTs and vaginal pH and Nugent score.
• Demographic associations: Facilitates analysis of CST associations with participant demographics such as race and age.
• Standardization for comparability: Provides a standardized classification approach to support reproducible between-study comparisons and meta-analyses.

Scientific Applications:

• CST assignment: Classifying vaginal microbiome samples into Community State Types for microbiome profiling studies.
• Epidemiological analysis: Investigating associations between CSTs and clinical metrics such as vaginal pH and Nugent score.
• Demographic studies: Assessing relationships between CST distributions and participant demographics including race and age.
• Cross-study integration: Enabling between-study comparisons and meta-analyses of vaginal microbiota datasets collected with different sequencing strategies and pipelines.

Methodology:

Classification is performed by computing similarity between sample taxonomic profiles and predefined reference centroids derived from 13,160 taxonomic profiles (1,975 women) and assigning samples to the nearest centroid using a nearest-centroid algorithm.