[Application of neural network autoencoder algorithm in the cancer informatics research]

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Cancer is a heterogeneous disease requiring early diagnosis and prognosis.
• Big data in biomedicine presents challenges for efficient data mining.
• Autoencoders offer unsupervised learning for feature extraction from complex data.

Purpose of the Study:

• To introduce the autoencoder model's structure and workflow.
• To summarize autoencoder applications in cancer informatics.
• To discuss future challenges and perspectives of autoencoders in this field.

Main Methods:

• Review of autoencoder architecture and unsupervised learning principles.
• Synthesis of existing literature on autoencoder applications in cancer informatics.
• Analysis of diverse biomedical data types utilized with autoencoders.

Main Results:

• Autoencoders demonstrate efficacy in learning features from biomedical data.
• Applications span various cancer types, integrating diverse data modalities.
• The review highlights successful implementations in cancer informatics.

Conclusions:

• Autoencoders are powerful tools for big data analysis in cancer research.
• Their unsupervised learning capability aids in feature extraction and data integration.
• Further development is needed to address challenges and expand applications in cancer informatics.