exploring proteins

[reimand0]

The page https://activedriverdb.org/protein/browse/ has a little too much information and should be compressed. I actually wonder if this type of a view is even necessary given that we have top 100 disease genes in a view.

1. we should only show the main isoform per gene and hide others behind the + symbol.
2. we should hide proteins with 0 PTM sites and/or 0 PTM mutations. If you think it is necessary, we can toggle that filter with a checkbox.
3. is it possible to rank by # PTM mutations?
4. Not clear what # mutations means here. is it all, cancer, population, disease? Should we define a set of radio buttons to show exactly what mutations are these?
5. We should add a little text under the Proteins title to tell the user what this page does.

[krassowski]

1. If done it would be exactly the same as https://activedriverdb.org/gene/browse/
2. All right, I will add a checkbox allowing to exclude non-PTM proteins, also to https://activedriverdb.org/gene/browse/
3. Yes, it is (as in https://activedriverdb.org/gene/browse/). It might be a little bit slow (in gene browse I do this but only for preferred isoforms) and require some creative SQL query but it is possible. Currently proteins are fast to load mainly because #PTM mutations are calculated on display. It is fast to do this for 25 proteins and still possible to do for 19017 genes. It will be harder to to this for 39159 proteins (isoforms). We may consider using some more advanced database which offers views functionality (or easily precompute* everything but it will make maintenance more complicated in future).
4. Those are all mutations. We have a filter already on https://activedriverdb.org/gene/browse/. Should I add this here too?
5. I added editable box into to below Proteins title so we can enter any text here. It is only visible after logging in into admin account. I added some more of such boxes in other places.

Overall, I feel that we already had this conversation, but with gene instead of protein ;)

Nevertheless there are still some problems: the gene/browse page is terribly slow now. It wasn't the last time I was checking it out. I will investigate this; maybe import of MC3 mutations is responsible, maybe it's some regression. I have an idea how to speed this up anyway.

When it comes to sole existence of the /protein/browse page: I think that we may want to remove it, though it could be useful for some users, and keeping it does not cost anything.

Please, let me know what you think.

*We already precompute if a mutation is PTM related or not (at import), now we would want to precompute sums of such mutations for each protein for possibly all datasets (and maybe site types?).

[reimand0]

Looking at this again - I think "explore all genes" and "explore all proteins" are two redundant views. We should keep one and remove the other, perhaps by viewing main isoform by default and making a checkbox "show all isoforms".

Regarding (4) above: let's use radio buttons instead of dropdown menus. Then all options will be visible immediately. There are four each (five with "all" option) so this should not be a major burden.

[reimand0]

Browsing these genes page by page still feels a little arbitrary because it is ultimately an alphabet soup and some numbers (no biologist will know all these genes by heart). Perhaps these two additions may help:

1. add short description of gene.
2. sort them by some heuristic so the "best ones" appear towards the top. We could use Marta's scores from ActiveDriver for cancer and/or Clinvar to make this a little more attractive for first browsing. Maybe even some weighted summary score of the two?

[reimand0]

Point two from above may actually improve our browsing of pathways as well. I will comment separately under the pathways view.

[krassowski]

Looking at this again - I think "explore all genes" and "explore all proteins" are two redundant views. We should keep one and remove the other, perhaps by viewing main isoform by default and making a checkbox "show all isoforms".

All right, I will retire the table view for browsing proteins. So let's assume all the points relate to gene browsing:

we should only show the main isoform per gene and hide others behind the + symbol.

Done (already in)

we should hide proteins with 0 PTM sites and/or 0 PTM mutations. If you think it is necessary, we can toggle that filter with a checkbox.

Done.

is it possible to rank by # PTM mutations?

Ye, it is.

Not clear what # mutations means here. is it all, cancer, population, disease? Should we define a set of radio buttons to show exactly what mutations are these?

Done (filters were already in, recently converted to radio buttons).

We should add a little text under the Proteins title to tell the user what this page does.

An editable box available under "Genes" title, visible after login.

1. add short description of gene.

Done

2. sort them by some heuristic so the "best ones" appear towards the top. We could use Marta's scores from ActiveDriver for cancer and/or Clinvar to make this a little more attractive for first browsing. Maybe even some weighted summary score of the two?

What would be the difference to gene lists we already have?