Integrating Expression Data-Based Deep Neural Network Models with Biological Networks to Identify Regulatory Modules...

Area of Science:

• Bioinformatics
• Genomics
• Oncology

Background:

• Lung adenocarcinoma is the most common lung cancer subtype.
• Regulatory mechanisms driving lung adenocarcinoma initiation are not fully understood.
• Identifying these mechanisms is a key area in bioinformatics research.

Purpose of the Study:

• To identify regulatory modules involved in lung adenocarcinoma carcinogenesis.
• To elucidate the roles of mRNAs, lncRNAs, and miRNAs in lung adenocarcinoma.
• To leverage deep neural networks for uncovering complex regulatory relationships.

Main Methods:

• Differential gene expression analysis of tumor versus non-tumor tissues.
• Construction and optimization of deep neural network (DNN) models.
• Identification of key RNAs and competing endogenous RNA (ceRNA) networks.

Main Results:

• Three distinct regulatory modules involving miRNAs, mRNAs, and lncRNAs were identified.
• These modules demonstrated consistent regulatory directions in lung adenocarcinoma.
• Validation in an independent dataset confirmed the modules' effectiveness.

Conclusions:

• The identified regulatory modules provide a reference for understanding lung adenocarcinoma carcinogenesis.
• This study highlights the potential of DNNs in identifying cancer-related regulatory networks.
• The findings contribute to the comprehension of ceRNA crosstalk in lung cancer.