SSR_VibraProfiler

SSR_VibraProfiler classifies plant varieties using simple sequence repeats (SSRs) extracted from next-generation sequencing data by constructing a DNA fingerprint database and applying dimensionality reduction and clustering for variety prediction.

Key Features:

• SSR Extraction: Extracts SSR numerical characteristics from sequencing data, treating the presence or absence of each SSR as a numerical characteristic while ignoring specific loci, flanking sequences, and occurrence counts.
• SSR Selection: Measures variance of numerical characteristics within and among varieties using a formula, and selects SSRs that exhibit intra-variety specificity and inter-variety polymorphism to form a 0,1 matrix.
• Dimensionality Reduction: Uses t-SNE (t-distributed Stochastic Neighbor Embedding) to project the SSR matrix onto a two-dimensional plane.
• Clustering: Applies K-means clustering to group individuals in the reduced-dimensional space.
• Classification Assessment: Preliminarily assesses classification performance by comparing cluster labels with true labels to evaluate effectiveness in variety detection.
• Variety Prediction: Constructs a recognition model based on the SSR matrix and applies it for variety identification, validated using leave-one-out cross-validation.
• DNA Fingerprint Database Construction: Encapsulates the entire process into a package for constructing an SSR variety DNA fingerprint database.

Scientific Applications:

• Plant Variety Identification: Distinguishes and predicts individual plant varieties without requiring a reference genome.
• Variety Detection and Protection: Contributes to the development, identification, and protection of new plant varieties.
• Molecular Marker Development: Provides a computational alternative to experimental methods for developing SSR molecular markers from sequencing data.

Methodology:

SSR_VibraProfiler extracts SSR presence/absence as numerical characteristics from sequencing data, measures their variance within and among varieties to select informative SSRs forming a 0,1 matrix, projects this matrix using t-SNE, performs K-means clustering, and constructs a recognition model for variety identification. The method was tested on a Rhododendron dataset of 40 individuals from 8 varieties.