site phenotype tool

[Bernadetadad]

@jbloom sometimes I just wish I could type in a mutation (e.g. P174Q) or site (174) and just get a print out of all phenotype values we measure for this site. I know we can find this all in heatmaps (an sometimes it's nice to see effects in the context of other mutations) and I could just search the summary.csv file for this but sometimes it's honestly just a pain to scroll though these heatmaps when I'm honestly just interested in effect of a single mutation. Could we incorporate something into the pipeline that just does that – if I type P174Q it prints effects of this mutation on all phenotypes we measured? I think many people outside our lab would also often would only be interested in effects on specific mutation and currently it take too many steps from people outside our group to get to that point unless they are already familiar with visualizations our group uses. Maybe @WillHannon-MCB could make such too and it could be incorporated into dms-vep pipeline? This honestly just needs to be some interactive filter for summary.csv spreadsheet the pipeline already creates.

[jbloom]

I believe what we probably really want here is something more general. I think we want a tool that would have the following properties, and should probably be built as some separate tool that then could be used for this but also more general tasks. @WillHannon-MCB, could something like this be built (or maybe it already exists):

• takes as input a CSV file with arbitrary columns
• you specify selections for one or more columns. For instance, site = 174 and amino acid = Q.
• The tool then does two things:
  1. display the CSV rows that just fit those identifiers (eg, look up the values)
  2. for columns specified as measurement columns, plots the full distribution of values in each column plus the value(s) of the selected sites. I think this is important, because often the numerical values are more informative if they can be contextualized relative to the full distribution.
• Maybe optionally you can also specify additional filter columns (eg, times seen >= 2) and only values that exceed that are plotted in the distributions.

Something like this would be a general tool that would be useful for a lot of what we do.

[WillHannon-MCB]

@jbloom, I can think of a couple of reasonable solutions to this. Are you envisioning something like a simple web app?

[WillHannon-MCB]

I made a simple widget in Observable with the functionality outlined by @jbloom. Unfortunately, this solution can't be automated or integrated into the pipeline.

I will add a link in the documentation if this solution seems reasonable. I'll also investigate how to make a similar widget that automatically gets embedded into the documentation by the pipeline.

[Bernadetadad]

@WillHannon-MCB the above works for me the only other thing that would be useful is to be able to display more than one mutation, just if say I want to make a table with effects of 5 different mutations (now it's particularly relevant to some H5 stuff). I think you can only have one mutation displayed at the moment due to how filter option works.

[WillHannon-MCB]

I can add that!

I'll make a better interface for selecting mutations so that you can display multiple at the same time. I'll probably make the format for selecting mutations a comma separated list that you type in. Does that work?

[Bernadetadad]

seems reasonable

[WillHannon-MCB]

@Bernadetadad, what do you think about this?

https://observablehq.com/@willhannon-mcb/dms-vep-mutation-search-tool

[Bernadetadad]

This is good, but would it be possible to display full column names? Ideally by splitting them in new lines and not widening the column as it's already quite a wide table.

[WillHannon-MCB]

Unfortunately, no.

[Bernadetadad]

unfortunate but I think this is fine, it does what I wanted.