ANOVAG3: A Hybrid Algorithm for Inferring Gene Regulatory Network Using Time Series Gene Expression Data

Abstract

GENIE3 achieves best results in inferring Gene Regulatory Network (GRN) with DREAM4 challenge data. Whereas, correlation coefficient derived from two-way analysis of variance (ANOVA) records best result for DREAM5 challenge data. Here we try to improve results of GENIE3 on time series gene expression data by using one-way ANOVA along time axis as a prior step to GENIE3. GENIE3 takes long time with huge number of genes so one-way ANOVA finds significant genes before execution of GENIE3. Integration between one-way ANOVA and GENIE3 is a hybrid algorithm entitled ANOVAG3. ANOVAG3 is applied only on time series gene expressions and takes less running time than GENIE3 with huge data. ANOVAG3 is compared with other algorithms which infer GRN by Area Under the Receiver Operating Characteristic Curve (AUROC) using DREAM5 challenge networks. Although ANOVAG3 is not dependent on perturbation data or transcription factors, it records comparable results for networks 1 and 3 and records best results for network 4 (AUROC =0.5628) of DREAM5 challenge data. ANOVAG3 records better results in DREAM 5 networks 2, 3 and 4 (AUROC= 0.5190, 0.6458 and 0.5628) compared to GENIE3 and PLSNET considering large scale time series data employed in this work.

Prediction of Vessel Traffic Volume in Ports Based on Improved Fuzzy Neural Network

“…Several approaches are often applied to predict the VTV, including, regression analysis prediction method (RAPM) [6], time series prediction method (TSPM) [7], [8], and gray theory prediction method (GTPM) [9], [10]. Among them, the RAPM is very sensitive to the selection of prediction variables and data.…”