Machine learning differentiates enzymatic and non-enzymatic metals in proteins

Abstract

Metalloenzymes are 40% of all enzymes and can perform all seven classes of enzyme reactions. Because of the physicochemical similarities between the active sites of metalloenzymes and inactive metal binding sites, it is challenging to differentiate between them. Yet distinguishing these two classes is critical for the identification of both native and designed enzymes. Because of similarities between catalytic and non-catalytic metal binding sites, finding physicochemical features that distinguish these two types of metal sites can indicate aspects that are critical to enzyme function. In this work, we develop the largest structural dataset of enzymatic and non-enzymatic metalloprotein sites to date. We then use a decision-tree ensemble machine learning model to classify metals bound to proteins as enzymatic or non-enzymatic with 92.2% precision and 90.1% recall. Our model scores electrostatic and pocket lining features as more important than pocket volume, despite the fact that volume is the most quantitatively different feature between enzyme and non-enzymatic sites. Finally, we find our model has overall better performance in a side-to-side comparison against other methods that differentiate enzymatic from non-enzymatic sequences. We anticipate that our model’s ability to correctly identify which metal sites are responsible for enzymatic activity could enable identification of new enzymatic mechanisms and de novo enzyme design.

Structural Bioinformatics and Deep Learning of Metalloproteins: Recent Advances and Applications

“…Structure-based predictions by DeepCys are only applicable to structures deposited in the PDB. MAHOMES is a recently developed approach aimed at distinguishing between enzymatic and non-enzymatic metals in MPs [84]. In this work, the authors applied fourteen different machine learning methods, including a neural network approach.…”

“…An "indirect" use of artificial intelligence in the study of MPs is the exploitation of AlphaFold [21,85] or RoseTTAFold [22] predictions to model or predict the occurrence of MBSs. In fact, the structural models in the AlphaFold database do not contain chemical MAHOMES is a recently developed approach aimed at distinguishing between enzymatic and non-enzymatic metals in MPs [84]. In this work, the authors applied fourteen different machine learning methods, including a neural network approach.…”

Learning to Identify Physiological and Adventitious Metal-Binding Sites in the Three-Dimensional Structures of Proteins by Following the Hints of a Deep Neural Network

“…By construction, each cluster contains MBSs with a specific metal ion, i.e., metal-substituted sites are assigned to distinct clusters. This clustering procedure, which is similar to what is done in ref ( 30 ), allows redundancy to be removed from the dataset (with the exception of proteins having multiple MBS, as described below).…”

Accurate prediction of transition metal ion location via deep learning

“…Structure based deep learning approaches have been used in the field of protein research for a variety of applications such as protein structure prediction, 28 prediction of identity of masked residues [30][31][32] , functional site prediction, 33,34 for ranking of docking poses, 35,36 prediction of the location of ligands, [36][37][38][39][40] and prediction of effects of mutations for stability and disease. 4 Current state of the art predictors for metal location are MIB, 23,42 which combines structural and sequence information in the "Fragment Transformation Method" to search for homologous sites in its database, and BioMetAll, 26 a geometrical predictor based on backbone preorganization.…”