ABSTRACT
Gene expression datasets provide powerful insight into functional genomics in life sciences. We used the maize Nested Association Mapping (NAM) expression datasets from maize qTeller and performed the Support Vector Machine (SVM) to classify the high-dimensional information on gene expression based on organ-specific characteristics: apex, ear, root, shoot, and tassel. We conduct a filtering process by removing null values and an ANOVA test to reduce the data complexity before the SVM analysis. We used the ratio of 70:30% for training and testing datasets and the cost value parameter equal to 0.1. We evaluate the SVM prediction using accuracy, precision, and recall functions. As a result, the accuracy rate is 100% for apex, ear, and tassel, while it is 88.89% for root and shoot with an Area Under the Curve (AUC) value = 0.9895. We obtained 8,470 gene expressions with the SVM weights and visualized the expression of the gene based on a weight value of ≥ 0.03. Hence, we found genes that are probably the key players in a specific metabolic pathway in maize organs. Moreover, the SVM provides new insight to analyze the gene expression datasets.
