Machine learning alternative to systems biology should not solely depend on data

Area of Science:

• Systems Biology
• Synthetic Biology
• Artificial Intelligence (AI) / Machine Learning (ML)

Background:

• AI/ML is increasingly proposed as an alternative to systems biology for understanding biological phenomena and synthetic biology design.
• Despite AI/ML's disruptive potential and successes, fundamental and practical challenges exist, particularly for analyzing chaotic or stochastic biological systems.

Purpose of the Study:

• To highlight potential pitfalls in applying AI/ML to biological research without careful consideration of data quality and suitability.
• To advocate for a balanced approach, integrating AI/ML with systems biology where appropriate.

Main Methods:

• Conceptual analysis of AI/ML applications in biology.
• Comparison of AI/ML with systems biology approaches.
• Discussion of data requirements and pre-processing for AI/ML in biological contexts.

Main Results:

• AI/ML, while powerful, carries risks of 'AI winters' if not implemented thoughtfully, similar to past issues in other fields.
• Emergent behaviors in chaotic or stochastic biological systems pose specific challenges for current AI/ML methodologies.

Conclusions:

• A proactive assessment of AI/ML suitability and data preprocessing is crucial for the research community.
• The future likely involves a hybrid approach, combining AI/ML with systems biology for robust biological discovery and design.