HiTaC

HiTaC performs hierarchical taxonomic classification of fungal internal transcribed spacer (ITS) sequences to improve accuracy in fungal identification and diversity assessment.

Key Features:

• Hierarchical Model Approach: Integrates the taxonomic tree hierarchy into machine learning models to inform classification decisions across ranks.
• Robustness with Limited Data: Maintains classification performance when training data are scarce.
• Handling Imbalanced Datasets: Manages class imbalance in sequence classification tasks.
• Performance on Noisy Data: Produces higher F1-scores and sensitivity across taxonomic ranks in the presence of noisy sequences.
• Versatility Across Sequence Variability: Accurately classifies ITS sequences of varying lengths and when identity differences exist between training and test datasets.

Scientific Applications:

• Diversity Estimation: Enables more accurate estimation of fungal diversity in ecological studies using ITS sequence data.
• Environmental Community Dynamics: Supports analysis of fungal abundance and distribution to study environmental community structure and change.
• Health and Disease Correlation Studies: Facilitates examination of correlations between fungal species abundance and health-related conditions, including infections.

Methodology:

HiTaC employs hierarchical machine learning models that incorporate the taxonomic tree structure during training and was evaluated using the TAXXI benchmark.