ANN-Glycolysis-Flux-Prediction

ANN-Glycolysis-Flux-Prediction predicts metabolic flux through the upper part of glycolysis using artificial neural networks to model NADH consumption and optimize enzyme concentrations in multienzyme cascade reactions.

Key Features:

• Artificial Neural Network Implementation: Employs ANN algorithms implemented with the neuralnet R package using "logistic" and "tanh" activation functions to capture nonlinear relationships.
• Input Data: Trains on experimental NADH consumption rates as the primary input signal for flux prediction.
• Enzyme Concentration Analysis: Uses concentrations of phosphoglucoisomerase, phosphofructokinase, fructose-bisphosphate-aldolase, and triose-phosphate-isomerase as predictive features.
• Performance Metrics: Reports RMSE values of 0.847 (logistic) and 0.804 (tanh) and R-squared values of 0.93 (logistic) and 0.94 (tanh).
• Validation: Assesses model reliability using cross-validation procedures.

Scientific Applications:

• Metabolic Engineering: Predicts flux distributions to support design and optimization of engineered glycolytic pathways.
• Bioprocess Optimization: Informs optimization of biotechnological processes reliant on glycolytic flux, such as fermentation or biosynthesis.
• Academic Research: Provides a computational framework for studying regulation and dynamics of upper glycolysis without exhaustive laboratory assays.

Methodology:

Collects experimental NADH consumption rates under varying enzyme concentrations; trains ANN models using the neuralnet R package with logistic and tanh activation functions; evaluates model accuracy using cross-validation.