MADS-Box Gene Classification in Angiosperms by Clustering and Machine Learning Approaches

Area of Science:

• Plant developmental biology
• Molecular genetics
• Bioinformatics

Background:

• MADS-box genes are crucial transcription factors regulating floral organogenesis.
• The ABCDE model explains floral identity but has limitations, especially for orchids.
• Existing classification methods like phylogenetic analysis can be error-prone and inefficient.

Purpose of the Study:

• To develop an advanced MADS-box gene classification system for angiosperms.
• To improve the accuracy and efficiency of classifying MADS-box genes involved in floral development.
• To address limitations of current models and methods, particularly for species like orchids.

Main Methods:

• A two-stage classification approach was developed.
• Curated reference datasets for eight MADS-box gene classes (A, AGL6, B12, B34, BPI, C, D, E) were clustered using phylogenetic analysis and unsupervised learning.
• Support vector machines with feature selection based on sequence similarity and MADS-box gene domain characteristics were employed, with a local BLAST model showing superior accuracy.

Main Results:

• The developed system achieved 93.3% accuracy in classifying *Phalaenopsis aphrodite* MADS-box genes, surpassing the previous iMADS tool (86.7%).
• The local BLAST model demonstrated higher accuracy compared to BindN and COILS features.
• The system identified and corrected classification errors made by traditional phylogenetic tree analysis.

Conclusions:

• The new MADS-box gene classification system is reliable and efficient for angiosperms.
• This advancement aids in a deeper understanding of floral organogenesis across diverse plant species.
• The system offers a robust alternative to phylogenetic tree analysis for large-scale MADS-box gene classification.