Graph embedding and geometric deep learning relevance to network biology and structural chemistry

Area of Science:

• Biological science
• Systems biology
• Network biology

Background:

• Graphs model complex biological relationships, crucial in systems biology since the early 2000s.
• Artificial intelligence (AI) techniques are increasingly applied to biological networks for tasks like classification and link prediction.
• Traditional machine learning methods struggle with large, dense biological networks due to computational demands and non-Euclidean geometry.

Purpose of the Study:

• To provide a comprehensive summary of main graph embedding algorithms.
• To highlight the potential of graph embedding in network biology.
• To discuss AI learning techniques in the context of geometric deep learning.

Main Methods:

• Review of graph embedding algorithms.
• Analysis of AI techniques applied to network biology.
• Exploration of geometric deep learning approaches.

Main Results:

• Graph embedding emerges as a powerful learning paradigm for biological network analysis.
• It facilitates complex data mining tasks by learning informative vector representations of graph data.
• Enables the use of efficient, non-iterative traditional models for tasks like classification and link prediction.

Conclusions:

• Graph embedding offers a promising solution for overcoming the limitations of traditional machine learning in network biology.
• The flourishing research in graph embedding is driven by its potential to unlock insights from complex biological networks.
• This review synthesizes current advancements, emphasizing the role of geometric deep learning.