GAEM: Genetic Algorithm based Expectation-Maximization for inferring Gene Regulatory Networks from incomplete data

In Bioinformatics, inferring the structure of a Gene Regulatory Network (GRN) from incomplete gene expression data is a difficult task. One popular method for inferring the structure GRNs is to apply the Path Consistency Algorithm based on Conditional Mutual Information (PCA-CMI). Although PCA-CMI e...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Computers in biology and medicine Jg. 183; S. 109238
Hauptverfasser: Niloofar, Parisa, Aghdam, Rosa, Eslahchi, Changiz
Format: Journal Article
Sprache:Englisch
Veröffentlicht: United States Elsevier Ltd 01.12.2024
Elsevier Limited
Schlagworte:
Algorithms
Bayesian network
Bioinformatics
Computational Biology - methods
Conditional Mutual Information
Convergence
Databases, Genetic
Datasets
Expectation-Maximization
Gene expression
Gene Expression Profiling - methods
Gene Regulatory Network
Gene Regulatory Networks - genetics
Genetic algorithm
Genetic algorithms
Humans
Internal Medicine
Machine learning
Maximization
Missing values
Models, Genetic
Optimization
Other
Conditional Mutual Information
Gene Regulatory Network
Genetic algorithm
Bayesian network
Expectation-Maximization
Missing values
ISSN:0010-4825, 1879-0534, 1879-0534
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In Bioinformatics, inferring the structure of a Gene Regulatory Network (GRN) from incomplete gene expression data is a difficult task. One popular method for inferring the structure GRNs is to apply the Path Consistency Algorithm based on Conditional Mutual Information (PCA-CMI). Although PCA-CMI excels at extracting GRN skeletons, it struggles with missing values in datasets. As a result, applying PCA-CMI to infer GRNs, necessitates a preprocessing method for data imputation. In this paper, we present the GAEM algorithm, which uses an iterative approach based on a combination of Genetic Algorithm and Expectation-Maximization to infer the structure of GRN from incomplete gene expression datasets. GAEM learns the GRN structure from the incomplete dataset via an algorithm that iteratively updates the imputed values based on the learnt GRN until the convergence criteria are met. We evaluate the performance of this algorithm under various missingness mechanisms (ignorable and nonignorable) and percentages (5%, 15%, and 40%). The traditional approach to handling missing values in gene expression datasets involves estimating them first and then constructing the GRN. However, our methodology differs in that both missing values and the GRN are updated iteratively until convergence. Results from the DREAM3 dataset demonstrate that the GAEM algorithm appears to be a more reliable method overall, especially for smaller network sizes, GAEM outperforms methods where the incomplete dataset is imputed first, followed by learning the GRN structure from the imputed data. We have implemented the GAEM algorithm within the GAEM R package, which is accessible at the following GitHub repository: https://github.com/parniSDU/GAEM. •GAEM learns GRN from incomplete data updating the imputed values using the GRN.•GAEM uses BNs to learn GRNs and applies GRNs’ information to impute missing values.•GAEM is not limited to a specific structure learning algorithm.•GAEM’s performance is notable in NMAR mechanisms when the network size is small.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:0010-4825
1879-0534
1879-0534
DOI:10.1016/j.compbiomed.2024.109238