Usage of metaPrDOS


Prediction Submission

Input query amino acid sequence

Input protein amino acid sequences in plain text or FASTA format into the form. Multiple FASTA formatted inputs are not acceptable. The server accepts the 20 single letter codes for standard amino acids. The code 'X' generally used for non-standard amino acids can not be accepted. Too long sequences with lengths greater than 500 residues are not accepted because of the limitation of predictors.

The single letter standard amino acid code
Amino acidSingle letter code
AlanineA
CysteineC
Aspartic AcidD
Glutamic AcidE
PhenylalanineF
GlycineG
HistidineH
IsoleucineI
LysineK
LeucineL
MethionineM
AsparagineN
ProlineP
GlutamineQ
ArginineR
SerineS
ThreonineT
ValineV
TryptophanW
TyrosineY

Prediction time

Although it depends on the length of query protein and condition of each component server, metaPrDOS usually takes from 5 to 10 minutes to predict a protein sequence. Occasionally, some external servers would not reply to our forwarded query within an hour (predefined limit in metaPrDOS server), because they may be down temporally or they may set a limit to handle a request in a single day from the same IP address. At that time, metaPrDOS will make a prediction using all available prediction results. For the purpose, we prepared trained meta-predictors for all combination of predictor in advance.


Output format of prediction results

E-mail outputs

Prediction results returned by e-mail includes followings: The prediction results are composed of 4 columns. The first column is a residue number, and the second is an amino acid type of the residue. If '*' in the third column, the residue is predicted disordered. Final column shows disorder probability of the residue.
Example of e-mail outputs
 No  AA Pred Probability
--------------------------
   1  A  *   0.85
   2  W  *   0.76
   3  L  *   0.82
   4  E  *   0.84
   5  A  *   0.82
   6  Q  *   0.79
   7  E  *   0.80
   8  E  *   0.79
   9  E  *   0.75
  10  E      0.68
  11  V      0.61
  12  G      0.53
--------------------------

HTML formatted outputs

The prediction result page is consisted of following three parts.
Example of HTML formatted outputs
HTML output

The prediction result of the two-state prediction (disorder/order)

The red residues are predicted to be disordered at the given prediction false positive rate. The user may change prediction false positive rate in this part. There is trade-off between prediction sensitivity (true positive rate) and false positive rate. If you permit higher false positive rate, you can obtain more sensitive prediction. The acceptable false positive rate or expected sensitivity depends on the aim of predictions. The following receiver-operating characteristic (ROC) curve shows the sensitivity (true positive rate) for a particular false positive rate. If the user want to recover at least 50% of disordered regions, the user should set false positive threshold at 2%. The default false positive rate is set at 5%.
ROC curves of metaPrDOS
ROC curves

Plot of disorder tendency

The middle part shows the plot of disorder tendency of each residue along the sequences. Residues beyond the red threshold line in this plot are predicted to be disordered. The user can change the size of the plot through the web-interface and also check the prediction results of each external prediction methods.

Combination of component predictors

Occasionally, some external servers did not reply to our forwarded query within an hour (predefined limit in the metaPrDOS server) because they may have been down temporarily or they may set a limit to handle requests in a single day from the same IP address. At that time, metaPrDOS will make a prediction using all available prediction results. The prediction method used for meta prediction can be checked at this part. When the number of available servers was limited, the performance of the meta approach deteriorated. The performance of such cases can be checked at prediction performances for all combinations of each component predictor page.

Download prediction results

At the bottom part, the user can download the raw prediction results in CSV format or CASP format.