Efficient Ensemble Feature Selection Based Boolean Modelling for Genetic Network Inference
Conference Contribution ResearchOnline@JCUAbstract
The reconstruction of Gene Regulatory Networks (GRNs) is important in systems biology, because GRNs can provide insight into regulatory interactions between genes. Various computational methods have been developed for this task, but most have low computational efficiency. In this paper, we introduce an ensemble feature selection approach used with Boolean network modelling for efficient and accurate inference of GRNs. Using discretized microarray expression data, the regulatory genes for each target gene are inferred using an estimated multivariate mutual information-based feature selection method. To remove irrelevant features, pair-wise mutual information score-based thresholding is used, and gene-wise precision and dynamic accuracy-based stopping criteria is used for the determination of the maximum indegree of a target gene. Further inference of regulatory genes is performed by ReliefF, an instance-based feature ranking method. We also introduce a new Append function, to obtain a single optimal set of regulatory genes by combining the selected sets of genes from the MRMR and ReliefF based on performance evaluation criteria. Our previous research finding, a Pearson correlation coefficient based Boolean modelling approach is utilized in this research for the efficient observation of the optimal regulatory rules associated with target genes and the selected regulatory genes. Experiments, evaluating the proposed approach are ongoing. To date we have obtained improved results in terms of structural accuracy and efficiency.
Journal
N/A
Publication Name
CIBCB 2021: 18th IEEE International Conference in Computational Intelligence in Bioinformatics and Computational Biology
Volume
N/A
ISBN/ISSN
978-0-908026-67-8
Edition
N/A
Issue
N/A
Pages Count
2
Location
Melbourne, Australia
Publisher
Federation University Australia
Publisher Url
N/A
Publisher Location
Melbourne, VIC, Australia
Publish Date
N/A
Url
N/A
Date
N/A
EISSN
N/A
DOI
N/A