SETApp: A machine learning and image analysis based application to automate the sea urchin embryo test

Area of Science:

• Ecotoxicology
• Environmental Chemistry
• Bioassay Development

Background:

• Environmental monitoring faces challenges in identifying numerous xenobiotic compounds.
• Combining high-throughput bioassays with chemical analysis is effective but labor-intensive.
• Streamlining toxicant identification is crucial for routine environmental analysis.

Purpose of the Study:

• Develop a high-throughput screening method for automated quantification of sea urchin larvae size increase and malformation.
• Create a predictive expert system to aid in toxicant identification pipelines.
• Facilitate the application of sea urchin embryo tests in effect-directed analysis.

Main Methods:

• Utilized a training set of 242 images to calibrate larvae size and malformation.
• Developed and compared two classification models based on partial least squares discriminant analysis (PLS-DA).
• Implemented Hierarchical PLS-DA for larvae classification and validated its performance.

Main Results:

• Hierarchical PLS-DA demonstrated high proficiency in classifying sea urchin larvae.
• Achieved a prediction accuracy of 84% during the validation phase.
• Compiled the developed scripts into a user-friendly standalone application (SETApp).

Conclusions:

• The developed SETApp offers an efficient, automated solution for sea urchin embryo testing.
• The system aids in complex toxicant identification, as demonstrated in a wastewater treatment plant case study.
• This method significantly reduces the workload associated with effect-directed analysis in ecotoxicology.