Talk:Sequence-based predictions (PKU)

This is just a place to add criticism and or positive feedback so we Jonathan and Sebastian have a possibility to improve. Please note, that we wont be able to adress all matters immediately but we will address them eventually!

source: [1]

We also want to mention our slogan for this current teamwork

criticism

• There are several typos and some spelling mistakes still in the text, e.g. reagions, ver (instead of very), cant, ties, tree (instead of three) ... and so on. I found myself almost correcting some of those ;)

• The visual outputs of PsiPred are not discussed so might as well be left out

• The source of the Uniprot structure is DSSP, this should actually be very easy to find: http://www.uniprot.org/manual/helix

GOpet section

• The link within the table does not work, and the link in the text above is labelled gene-ontology.com although it is really AmiGO http://amigo.geneontology.org/cgi-bin/amigo/gp-assoc.cgi?gp=UniProtKB:P00439
• You state that 7 of the predicted annotations are found within the 10 from AmiGO .... but I can not reproduce this (e.g. 0003824 is not in there). The AmiGO GO terms should not be viewed as a gold standard anyway but you explained the go terms assignment very well in the text.

Transmembrane section

• As far as I see it no group did this so don't take it personally but Q9YDF8 really is quite a special case and investigation into what 'intramembrane' in Uniprot means and why there are differences in the annotation between PDBTM and OPM (which state the reason directly on the entries' page) might have shed some light into the problems the methods are facing with this structure and what is special about these channels in general
• Why show the output of PolyPhobius if you never talk about it?
• I have to disagree, it IS the goal of these predictions to recognize intramembrane as well as transmembrane regions (PolyPhobius is simply too old to know). I agree with the rest of your conclusion though

positive feedback

• The calculated scores e.g. for the secondary structure prediction are very helpful to asses the performance of the methods. The scores are additionally well displayed.

• Altogether the wiki entry is very interesting and nice to read.

• Good idea to compare SignalPv3 and v4 and only use the beginning parts of the sequence

suggestions

• You could consider removing either the first paragraph (In this task we will find out, how much information ...) or the task description and merging the contents as these two sections are a bit redundant.

• Some sentences seem a bit bewildering e.g. "Overall the results one can derive from several methods really are the same but one also can see the difference in the methods", but this may be just my perception.
• I understand your way of organising the scripts makes this a bit hard, but I strongly suggest a better seperation between protocol and evaluation
• try to adhere to a more scientific writing style, i.e. no 't tells you a lot and nothing' and so on
• You used many prediction methods for the disorder part (which is great) but did hardly any evaluation. If the methods really are all the same I would have preferred to just hear that and hide the data on an extra page for those who want to see for themselves