A Convolutional Deep Clustering Framework for Gene Expression Time Series.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Gene expression levels regulate cellular processes.
• Gene expression time series reveal cellular dynamics.
• Clustering gene expression patterns is crucial for high-throughput experiments.

Purpose of the Study:

• To propose DeepTrust, a novel deep learning framework for gene expression time series clustering.
• To overcome limitations of existing clustering techniques, such as handling short time series and preserving temporal structure.

Main Methods:

• Transforming gene expression time series data into images for richer representations.
• Applying a deep convolutional clustering algorithm on the generated images.
• Utilizing enrichment analyses to validate biological plausibility.

Main Results:

• DeepTrust demonstrates efficiency in clustering gene expression time series on simulated and biological datasets.
• The framework outperforms widely used clustering techniques.
• Detected clusters show biological plausibility through enrichment analyses.

Conclusions:

• DeepTrust offers an effective deep learning-based approach for gene expression time series clustering.
• The image transformation and deep convolutional clustering method enhances the analysis of temporal gene expression data.
• This framework provides valuable insights into cellular dynamics and gene function.