Volume 11, Issue 1 (6-2024)
jhbmi 2024, 11(1): 72-82 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Einipour A. Providing a Multi-Scale Self-Encoding Method to Improve Clustering and Analysis of Single-Cell Data. jhbmi 2024; 11 (1) :72-82
URL: http://jhbmi.ir/article-1-867-en.html
URL: http://jhbmi.ir/article-1-867-en.html
Department of Computer Engineering, Andimeshk Branch, Islamic Azad University, Andimeshk, Iran
Abstract: (839 Views)
Introduction: In bioinformatics, analyzing single-cell data is crucial for understanding cellular functions' complexities. However, this analysis faces challenges like inefficient dimensionality reduction and suboptimal clustering. This study aimed to present a method that enhances the clustering of single-cell data, improves reconstruction quality, and reduces data dimensions.
Method: This paper introduces SAMS (Single-cell Analysis using Multi-Scale Autoencoder), which uses a multi-scale autoencoder model to improve the challenges in single-cell data analysis. The SAMS method involves three primary steps: (1) data preprocessing and normalization, (2) employing a deep neural network model to reconstruct and reduce data dimensions with the help of a multiscale autoencoder, and (3) clustering the reduced data using the K-means algorithm to assess the method's performance.
Results: The SAMS method was implemented using Python on single-cell datasets. The results demonstrate that SAMS can effectively visualize cells in a two-dimensional space with an average Nearest Neighbor Error (NNE) rate of 89%, indicating a strong preservation of data structure. Additionally, the Silhouette index and Davis-Bouldin index, which measure clustering accuracy, show significant improvement with averages of 0.66 and 0.50, respectively.
Conclusion: The proposed SAMS method by combining the multiscale self-encoder model and the K-means algorithm could obtain better results than the previous methods. Its application in single-cell data analysis can aid researchers in gaining deeper insights into cellular functions and discovering new patterns.
Method: This paper introduces SAMS (Single-cell Analysis using Multi-Scale Autoencoder), which uses a multi-scale autoencoder model to improve the challenges in single-cell data analysis. The SAMS method involves three primary steps: (1) data preprocessing and normalization, (2) employing a deep neural network model to reconstruct and reduce data dimensions with the help of a multiscale autoencoder, and (3) clustering the reduced data using the K-means algorithm to assess the method's performance.
Results: The SAMS method was implemented using Python on single-cell datasets. The results demonstrate that SAMS can effectively visualize cells in a two-dimensional space with an average Nearest Neighbor Error (NNE) rate of 89%, indicating a strong preservation of data structure. Additionally, the Silhouette index and Davis-Bouldin index, which measure clustering accuracy, show significant improvement with averages of 0.66 and 0.50, respectively.
Conclusion: The proposed SAMS method by combining the multiscale self-encoder model and the K-means algorithm could obtain better results than the previous methods. Its application in single-cell data analysis can aid researchers in gaining deeper insights into cellular functions and discovering new patterns.
Type of Study: Original Article |
Subject:
Bioinformatics
Received: 2024/05/31 | Accepted: 2024/07/9
Received: 2024/05/31 | Accepted: 2024/07/9
Audio File [MP3 734 KB] (46 Download)
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
