Welcome to Metallopred

An Artificial Neural Network (NN) based 3 layer Prediction (or classification) tool for Metalloproteins through Protein Sequence.

For Metallopred application, refer standalone.

Metallopred Overview:

The large amount of proteomic data is available for a variety of organisms, allow researchers to efficiently identify novel proteins in distantly related organisms and annotating them. A faster means of annotation would be to match them with the already annotated sequences using sequence based similarity search method like BLAST. It is a discrete method of calculating the similarity between protein sequences simply by measuring the number of matches and mismatches. However, the function of a protein not only depends on its primary sequence but also on how the protein folds into 3D structure which in turn also depends on the hydrophobicity and hydrophilicity properties of the proteins. Therefore it is needed to capture sequence order information, short term and long term interactions between amino acids in a protein sequence as well as to capture proportion of hydrophobicity and hydrophilicity properties of the proteins in order to correctly annotate the raw protein sequence.

Therefore, we developed this tool for predicting the metal binding proteins from sequence derived features. These methods achieved good prediction accuracies and could nicely complement experimental approaches for identification of metal binding proteins. The prediction methods are unique in the sense that they do not require homologous protein sequences.

Metallopred Prediction:

We developed a tool consisting of 3 level of hierarchical classification using artificial neural network (ANN). First layer of classification decides whether protein sequence is Metal Ion Binding or Non-Metal Ion Binding. If the protein sequence is Metal Ion Binding, it is classified into either of major classes, Alkali Earth Metal Ion Binding, Alkali Metal Ion Binding and Transition Metal Ion Binding at second level of classification. In the third level of classification, the tool finally predicts the specificity of the protein to bind with a metal ion. Sequence derived features like physicochemical properties; amino acid composition and sequence based correlation of amino acids (pseudo amino acid) were used during the training, testing and validation of the tool. Our tool is robust and successfully classifies the novel protein sequence into metal binding protein, then into its major class and finally predicts specific metal binding.