0
$\begingroup$

I am new to the bioinformatics field. I have positive and negative protein sequences for acetylation PTM. Now, I want to train a classifier, say SVM. What will be the next step? How can I convert these sequences into usable features? Any information or links would help.

>P31327_55|1|testing
KAQTAHIVLEDGTKMKGYSFGHPSSVA
>P31327_57|1|testing
QTAHIVLEDGTKMKGYSFGHPSSVAGE
>P31327_119|1|testing
APDTTALDELGLSKYLESNGIKVSGLL
>P31327_157|1|testing
LATKSLGQWLQEEKVPAIYGVDTRMLT
$\endgroup$
2
  • 2
    $\begingroup$ It may be worth checking what other folk do for motif classification, for example see elm.eu.org/elms/candidates.html#MOD_acetylation which lists the different motifs (different enzymes recognise different motifs). Although ELM uses metadata such as protein localisation to filter the motifs for much better results. $\endgroup$ Mar 5 at 16:06
  • 1
    $\begingroup$ Q to others. a ML features (measurable property) is a different thing than the sequence feature present in a GenBank file (i.e. an annotation on the sequence, e.g. acetylation sites). This the second time I have seen this tag usage, is this something worth discussing in Meta? $\endgroup$ Mar 5 at 16:08
1
$\begingroup$

Just a recommendation: Be careful when making your data sets.

P31327_57, P31327_119, P31327_157 are all annotated as "By similarity" to the mouse entry in Swiss-Prot https://www.uniprot.org/uniprot/P31327#ptm_processing, which means there is no direct experimental evidence in the human entry. If you do have publications that report experimental evidence for these PTMs, please don't hesitate to share them with the UniProt team.

I also recommend reading this UniProt help page about using UniProt data in negative datasets if you haven't already done so: https://www.uniprot.org/help/negative_datasets

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.