MARSpred

Sequence Name: The user can give a name of sequence which he/she wish to input for the prediction.

Prediction Tool Options

1. Only MARSpred based prediction tool (Default): This tool requires protein sequence in standard FASTA format (User can submit/upload multiple protein sequences). MARSpred server will predict whether protein sequence belongs to mitochondrial or cytosolic AARSs. It will also give the SVM score in result section.

2. icaars and MARSpred based prediction tool: In this tool, we have implemented our previously developed web-server icaars (Panwar and Raghava 2010) with MARSpred. It requires only single protein sequence in FASTA format. Firstly, icaars server will predict that whether protein sequence is belongs to AARS or Non-AARS. If protein sequence is predicted as AARSs then MARSpred server will predict whether protein sequence belongs to mitochondrial or cytosolic AARSs.

Prediction Approach Options

1. Amino acid composition approach based prediction: In this percentage composition of all 20 amino acids were calculated, which inturn were used to derive the weight corrosponding to each amino acid. It was done by substracting the composition data. To determine the any unknown protein, compositions is calculated and then corrosponding weight is multiplied to it. All the 20 values determined in this way is summed up to get the cumulative score. If the cumulative score is less than 0 then it will be classified as Cytosolic Aminoacyl-tRNA synthetases protein and If the score is greater than 0 then it will be classified as Mitochondrial Aminoacyl-tRNA synthetases protein.

2. SAAC approach based prediction: In the split amino acid composition (SAAC) based approach, protein sequence devided into three parts (N-40, C-60 and Rest). All the 60 split amino acids were calculated, which inturn were used to derive the weight corrosponding to each split amino acid. To determine the any unknown protein,the split amino acid compositions is calculated and then corrosponding weight is multiplied to it. All the 60 values determined in this way is summed up to get the cumulative score. If the cumulative score is greater than 0 then it will be classified as Mitochondrial-AARS and vice-versa.

3. SA-SAAC approach based prediction: This is a appraoch of split amino acid composition of selected amino acids. We have selected 13 significant attributes by using WEKA 3.6.0 version. WEKA is a package of java programs for machine learning. We have used attribute evaluator for SVMAttributeEval (parameter -X 1 -Y 0 -Z 0 -P 1.0E-25 -T 1.0E-10 -C 1.0 -N 0) method with ranker (parameter -T -1.7976931348623157E308 -N -1). We have selected 4 (Asp, Glu, Lys, Arg), 1 (Ser) and 8 (Asp, Glu, His, Lys, Leu, Gln, Ser, Tyr) amino acids from N-termini, C-termini and intermediate regions respectively. In this percentage composition, protein sequence devided into three parts (N-40, C-60 and Rest) and 13 split amino acids compositions of selected amino acids were calculated, which inturn were used to derive the weight corrosponding to each selected split amino acid. To determine the any unknown protein,the split amino acid of 13 selected amino acids is calculated and then corrosponding weight is multiplied to it. All the 13 values determined in this way is summed up to get the cumulative score by SVM. If the cumulative score is greater than 0 then it will be classified as Mitochondrial-AARS and vice-versa. By using this appraoch we have achieved 98.33% sensitivity, 92.50% specificity, 96.00% accuracy with 0.92 MCC. Here, we have predicted 98.33% mitochondrial-AARS correctly (98.33% sensitivity).

SVM Threshold Options: The method is based on Support Vector Machine (SVM). Depending upon the threshold value which user choices, SVM will classify the unknown protein into Mitochondrial-AARS or Cytosolic-AARS protein. The default threshold is 0.0. If user want less sensitivity but more specificity, then higher threshold value should be specified, but if opposite is anticipated then lower threshold value should be choosen. So, the expected outcome will depends on the trade-off between sensitivity and specificity.

You can see and analyse our results - HERE