Intron identification approaches based on weighted features and fuzzy decision trees

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Current splice site prediction relies on the classical intron definition model (IDM).
• The computation-oriented IDM (CO-IDM) offers enhanced details for intron flanks of splice sites (IFSSs).

Purpose of the Study:

• To develop a novel approach using fuzzy decision trees (FDTs) for improved intron identification.
• To enhance the accuracy of computational predictions for splice sites.

Main Methods:

• Utilized weighted intronic sequence features (ISFs) including twelve uni-frame patterns (UFPs) and forty-five multi-frame patterns (MFPs).
• Employed gain ratios for improved intron identification performance.
• Fuzzified genomic sequence features using membership functions and unsupervised self-organizing map (SOM) technique.
• Generated interpretable fuzzy rules through global weighting and cross-referencing.

Main Results:

• Demonstrated significant improvement in intron identification accuracy.
• The proposed FDT method proved effective in enhancing predictive performance.
• Developed an online tool for inferring unknown genomic sequences as introns.

Conclusions:

• Fuzzy decision trees offer a powerful and interpretable method for intron identification.
• The novel approach enhances the accuracy of computational predictions in genomics.
• The developed online tool provides a practical resource for biological sequence analysis.