Competitive fitness analysis using Convolutional Neural Network.

Area of Science:

• Computational Biology
• Genetics
• Bioinformatics

Background:

• Competitive fitness assays are crucial for understanding organismal adaptation.
• Green fluorescent protein (GFP) markers are commonly used but can complicate analysis.
• Image-based analysis offers potential for high-throughput fitness measurements.

Purpose of the Study:

• To develop and validate a Convolutional Neural Network (CNN) model for estimating competitive fitness in Caenorhabditis elegans.
• To automate the counting of wild-type and GFP-expressing nematodes in fitness assays.
• To compare the accuracy and efficiency of the CNN model against traditional manual scoring methods.

Main Methods:

• Utilized Caenorhabditis elegans as a model organism for competitive fitness assays.
• Developed a Convolutional Neural Network (CNN) for image classification to distinguish wild-type from GFP-expressing nematodes.
• Quantified nematode populations by analyzing images and comparing model-derived counts with visual scoring.

Main Results:

• The CNN model achieved high performance metrics, with average precision ranging from 0.79 to 0.87 and average recall from 0.84 to 0.92, varying with nematode density.
• The automated counting process was at least 20 times faster than manual counting.
• The model demonstrated accuracy comparable to manual scoring, while eliminating human error.

Conclusions:

• The developed CNN-based procedure provides an efficient and accurate method for estimating competitive fitness in C. elegans.
• This image analysis tool has the potential to be optimized and applied to other image-based biological research areas.
• The model's availability on GitHub facilitates further development and adoption in the scientific community.