FAQ (Frequently Asked Questions)
1. I submitted the sequence, but still I did not get any email with results?There can be many reasons why it happens:
- check the Spam folder (in many cases automatically generated emails send by robots, scripts are marked as a spam)
- the second most common reason is simply, because the result is not ready yet.
2. The result are not ready yet, how long will it take?Do not expect results in seconds or minutes after sending query. Metadisorder uses over 20 programs to calculate final result, so it will take a while. Unfortunately, it depends also on remote servers (e.g. Phyre, Poodle-l, for details see "Method overview") so it means that sometimes you have to wait even days if for instance remote server is switch off, broken etc.
I strongly suggest to bookmark the result page after sending the query).
3. Why I cannot submit sequences longer than 1000 amino acids?There are two main reasons why we exclude sequences longer than 1000 amino acids:
- from technical reason: first our server uses third-party programs which not always can be installed locally (e.g. POODLE-S, POODLE-L, PHYRE), in most cases those methods have length limitation or/and their speed strongly depends on query length, second we also have to take into consideration that some of this algorithm has exponential complexity, which can be demanding even if those programs are installed locally (e.g. hhseach run over cdd database can take few minutes on current machine [16 CPU and 16GB RAM] and for longer sequences it simply crashes because of the lack of memory)
- although there are many proteins which have more than 1000 amino acids, in most cases functional proteins has less than 1000 amino acids, moreover manual splitting of longer proteins will not affect the final outcome, for example spliceosom protein PRPF8 which has 2335 residues, has only one domain solved by X-ray (1760-2016), the rest of the protein is disordered which is easily predicted by most of the predictors, even if you manually split it to 3 parts. This can be easily explained also by looking how the predictors works (not only for predicting disorder). They, usually use so called n-size windows for constructing features fed by machine learning methods (e.g. if method uses 11-size window, it means that when analyzing given amino acid it will take into consideration also 5 amino acids from the left and from the right of it, and use those 11 segment to create input features which will be send for instance to neural network layers). The remaining part of the sequence is not taken into consideration, so the prediction accuracy should be similar (or almost identical) both for one long sequence or sliced parts, which can be analyzed keeping the server restrain that is 1000 amino acids length
4. Other problems and sugestions?If you have other problems or suggestions, do not hesitate to contact directly server developer Lukasz Kozlowski