AI-driven parametrization of Michaelis–Menten maximal velocity: Advancing in silico new approach methodologies (NAMs)
Achilleas Karakoltzidis, Spyros P Karakitsios, Dimosthenis Andreas Sarigiannis · NAM journal. · 2025
The development of mechanistic systems biology models necessitates the utilization of numerous kinetic parameters once the enzymatic mode of action has been identified. Simultaneously, wet lab experimentation is associated with particularly high costs, does not adhere to principles of reducing the number of animal tests , and is a time-consuming procedure. Alternatively, an artificial intelligence-based method is proposed that utilizes enzyme amino acid structures as input data. This method combines NLP techniques with molecular fingerprints of the catalysed reaction to determine Michaelis–Menten maximal velocities ( V max ). The molecular fingerprints employed include RCDK standard fingerprints (1024 bits), MACCS keys (166 bits), PubChem fingerprints (881 bits), and E-States fingerprints (79 bits). These were integrated to produce reaction fingerprints. The data entries were sourced from SABIO RK, providing a concrete framework to support training procedures. After the data preprocessing stage, the dataset was randomly split into the training set (70 %), validation set (10 %), and test set (20 %) ensuring unique amino acid sequences for each subset. The data points with structures similar to the ones used to train the model as well as uncommon reactions were employed to further test the model. The developed models were optimized during the training procedure to predict V max values efficiently and reliably. Utilizing a fully connected neural network , these models can be applied to all organisms. Amino acid proportions of enzymes were also tested resulting in an unreliable predictor for the V max value. During testing, the model demonstrated better performance on known structures compared to unseen data. In the given use case, the model trained solely on enzyme representations achieved an R -squared of 0.45 on unseen data and 0.70 on known structures. When enzyme representations were integrated with RCDK fingerprints, the model achieved an R-squared of 0.46 on unseen data and 0.62 on known structures.