Regression plane concept for analysing continuous cellular processes with machine learning.

Area of Science:

• Computational Biology
• Genetics
• Machine Learning

Background:

• Biological processes are often continuous, but traditional methods discretize them, limiting discovery.
• Supervised machine learning typically relies on classification, which may not capture the nuances of continuous biological data.

Purpose of the Study:

• To introduce a user-friendly, class-free phenotypic supervised machine learning tool called the Regression Plane (RP).
• To enable the description and exploration of biological data in a continuous manner.
• To demonstrate the utility of RP in identifying genes, modeling complex processes, and characterizing continuous biological phenomena.

Main Methods:

• Developed and applied multi-parametric active regression to create the Regression Plane (RP).
• Compared traditional classification with regression in simulated experimental setups.
• Utilized RP for gene identification in triglyceride regulation and analysis of time-lapse mitosis data.

Main Results:

• RP enables class-free phenotypic supervised machine learning for continuous biological data exploration.
• Successfully identified genes involved in triglyceride level regulation in human cells.
• Demonstrated RP's capability to model complex biological processes, such as mitosis, at high resolution.
• Showcased the continuous nature of hemocyte differentiation in Drosophila melanogaster.

Conclusions:

• The Regression Plane (RP) offers a powerful approach for analyzing continuous biological data, overcoming limitations of discrete classification methods.
• RP facilitates deeper insights into gene function, complex dynamic processes, and developmental biology.
• This methodology advances phenotypic discovery by embracing the continuous nature of biological systems.