About Metallopred

Cited by:

Metallopred is an Artificial Neural Network (NN) based tool for prediction of protein sequences for binding to Metal ions. It predicts a 3 layer classification for Metalloproteins.

Artificial Neural Networks (NN) are mathematical or computational models which are based on structural and functional analogy to Natural Neural Networks and thus are capable of learning in the same fashion as brain (neurons) of living organisms learn.


Classification System of Metallopred

The classification system of Metallopred comprises of 3 levels and is inherited form some chemistry backgrounds and Metal Ion Binding category in Gene Ontology.



Parameter choices in Metallopred

Metallopred uses 3 types of parameters (a.k.a. Descriptors) and their combinations for generation or conversion of sequence to numerical parameters for feeding in Neural Network Architecture for it's prediction.

  1. Pseudo amino acid composition (37 parameters)
  2. Amino acid composition (20 parameters)
  3. Physicochemical properties (12 parameters)
  4. Pseudo amino acid + Amino acid composition (57 parameters)
  5. Pseudo amino acid + Physicochemical properties (49 parameters)
  6. Amino acid composition + Physicochemical properties (32 parameters)
  7. Pseudo amino acid + Amino acid composition + Physicochemical properties (69 parameters)

Description of the 3 parameter types:

  1. Pseudo amino acid composition: Comprises of 37 factors, 20 of which are amino acid compositions with a weighted factor and rest 17 are joint correlation of amino acids which describes the sequence of arrangement of amino acids in the protein sequence (described by K. C. Chou).
  2. Amino acid composition: Comprises of 20 factors representing composition of each standard amino acid in the sequence.
  3. Physicochemical properties: Comprises of 12 factors representing physicochemical properties of sequence, calculated by EMBOSS (EBI) package: Mol. Wt., Charge, Isoelectric pt., Mole percentages of Tiny, Small, Aliphatic, Aromatic, Non-polar, Polar, Charged, Acidic, Basic amino acid residues.


Neural Network System of Metallopred

Metallopred uses in total of 7 Neural Network Clusters according to parameter choice combinations and it's each NN Cluster comprises of 5 Neural Networks (in total of 7x5 = 35 Neural Networks) clubbed together for prediction of the given sequence into a 3 step heirarchy.

It's First level has 1 Neural Network Architecture as:

Second level has 1 Neural Network Architecture as:

Third level has 3 Neural Network Architectures as:

Datasets and Accuracy of Metallopred

For the purpose of training of Neural Networks we used protein sequence data from RCSB Protein Databank.

For detailed accuracy and datasets refer Downloads.