An Approach of Anomaly Detection and Neural Network Classifiers to Measure Cellulolytic Activity.

Area of Science:

• Biotechnology
• Microbiology
• Computational Biology

Background:

• Traditional methods for detecting cellulolytic microorganisms rely on subjective halo formation on solid media, lacking real-time monitoring capabilities.
• This subjectivity and lack of real-time analysis limit high-throughput screening of enzymatic activity.

Purpose of the Study:

• To develop a computational analysis method for visual patterns of cellulolytic activity using artificial neural networks.
• To enable objective and real-time monitoring of enzymatic hydrolysis.

Main Methods:

• Generated a data library of absorbance readings and RGB values from enzymatic hydrolysis using spectrophotometry and a prototype camera-based system (Enzyme Vision).
• Developed a linear regression model to predict absorbances from RGB color patterns.
• Trained, validated, and tested a neural network model to predict cellulolytic activity based on color patterns.

Main Results:

• Established six new descriptors for predicting temporal changes in enzymatic activity.
• The neural network model achieved regional classification of a cellulolytic microorganism halo into three of six learned classes.
• The computational method demonstrated agreement with spectrophotometry-based measurements.

Conclusions:

• The developed artificial neural network approach offers a viable alternative for high-throughput screening of enzymatic activity in real time.
• This method provides objective and quantitative analysis of cellulolytic activity, improving upon traditional subjective techniques.