Comparison of SVM and Spectral Embedding in Promoter Biobricks' Categorizing and Clustering

• DOI: 10.48550/arxiv.1902.05724
• Publication Date: 2019-01-01
• Authors: Shangjie Zou
• Keywords: Spectral Clustering; Discriminative model

Background: In organisms' genomes, promoters are short DNA sequences on the upstream of structural genes, with function controlling genes' transcription. Promoters can be roughly divided into two classes: constitutive and inducible promoters. clear functional annotations practical synthetic biology biobricks. Many statistical machine learning methods have been introduced to predict functions candidate Spectral Eigenmap has proved an effective clustering method classify biobricks, while support vector (SVM) is a powerful algorithm, especially when dataset small. Methods: The algorithms: spectral embedding SVM applied same 375 prokaryotic For embedding, Laplacian matrix built edit distance, followed by K-Means Clustering. represented numeric serve as for trainning. Results: achieved high predicting accuracy 93.07% in 10-fold cross validation classification promoters' transcriptional functions. eigenmap (spectral embedding) based editing distance may not capable extracting discriminative features this task.