Its2vec: Fungal Species Identification Using Sequence Embedding and Random Forest Classification.

Area of Science:

• Mycology
• Bioinformatics
• Computational Biology

Background:

• Fungal taxonomic classification is crucial for understanding biodiversity and ecological roles.
• Massive internal transcribed spacer (ITS) barcode data require efficient processing tools.
• High ITS sequence variation and computational demands challenge existing fungal identification methods.

Purpose of the Study:

• To develop a bioinformatics tool, Its2vec, for accurate fungal species-level classification using ITS barcodes.
• To address the challenges of high dimensionality and computational requirements in fungal ITS data analysis.

Main Methods:

• Assembled a comprehensive ITS database of over 25,000 fungal species.
• Utilized a word embedding algorithm for dimensionality reduction of ITS sequences into dense vectors.
• Developed a random forest classifier for species identification based on these vectors.

Main Results:

• Its2vec achieved accuracy comparable to state-of-the-art fungal species identification predictors.
• The Its2vec model demonstrated superior capability in handling large datasets and significant dimensionality reduction.
• Successfully classified fungal ITS barcodes to the species level.

Conclusions:

• Its2vec provides an efficient and accurate solution for fungal species identification from ITS barcode data.
• The tool is expected to aid in microbial community characterization and deepen the understanding of fungal functional mechanisms.
• Its2vec facilitates the study and preservation of fungal biodiversity by enabling large-scale analysis.