Help for mitpred server
The user can give a name of sequence which he wish to input for the prediction.
Depending upon the option which will be opted for prediction or number of queries in que, time required to serve the query will vary from 10 seconds to more than 5 minutes. Instead of waiting till the prediction is complete, this option provide oppurtunity to put the email-id of user to the server. As the prediction is complete, user will receive an email to intimate the same. Only alphabets or numerics or '_' should be there in email-id.
On the text box beside this option, sequence whose localization has is to be predicted, can be pasted. Option of uploading the sequence file in fasta format is also there. By default the server takes only single letter code of amino acids. The server also has the capability to ignore all the non-standard characters such as ,*%!@$% etc.
Prediction options:There are four methods of prediction:-
SVM: The second method is based on Support Vector Machine (SVM). Depending upon the threshold value which user choses, SVM will classify the unknown protein into mitochondrial or non-mitochondrial. The default threshold is 0.5. 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.
BLAST+SVM: In this both BLAST (BlastP) and SVM will be used for prediction. Dataset (both mitochondrial and non-mitochondrial proteins) used for training of SVM will be used as database and search will be performed against this database. Hit having maximum score and minimum e-value will be taken as significant hit. The threshold e-value upon which a hit is considered as significant is 1e-4. Depending upon the class to which this protein belongs, classification will be done. If no hit will be found then SVM will be used for prediction.
HMM based Pfam search+SVM: In this approach, by use of HMMER, HMM based Pfam search will be done to examine which domain(s) is/are present in the input sequence. If any domain is found then the class to which it belongs in our domain catalogue will be scaned. Our domain catalogue classify Pfam domains into three classes (i) those that found exclusively in mitochondiral proteins, (ii) those that found in proteins localized at sites other than mitochondria and (iii) those found both in mitochondrial and non-mitochondrial proteins. If input sequence contains even one class (i) or (ii) domain then it is directly assigned to that class directly. But if only class (iii) or no domain is found then SVM is used for prediction.
If you still have any doubt or got some suggestion then kindly contact us.