search

tbaccata / amica

amica: an interactive and user-friendly web-based platform for the analysis of proteomics data
GNU General Public License v3.0
22 stars 5 forks source link

requesting a new QC feature #4

Closed fazeliniah closed 2 years ago

fazeliniah commented 2 years ago

Dear Sebastian and amica team, Thanks again for all your work to make this useful tool publicly available. I understand that the imputation is necessary for some of the features in amica, however this may affect the analysis for some of the proteomics data. This includes the analysis of the IP samples, where the bait protein and some potential binders are only present in one group. Or the proteomic analysis of the samples with the knockout genes. Is there any options in amica to count for these cases? Can we have a new feature to generate new plots (maybe Venn?) for the samples before imputation? Thanks

tbaccata commented 2 years ago

Hello! Thank you again!

Barplots for the number of id. proteins per sample, and the percentage of missing values per samples are available in the QC tab, both are produced with the uploaded intensities before imputation:

qc_barplots

In the Analysis parameters you can specify to only include proteins that have been identified in at least n replicates of the same group (e.g 2 out of 3 replicates), and you can also apply this filtering on valid values filter on the bait samples only:

filter_on_valid_values

That being said, there are use cases were it would be beneficial to not perform the imputation at all. This feature would interfere with the current functionality though, heatmaps (and to some degree profile plots) can not deal with missing values. I'll need to think about it, how to include this feature without breaking anything.

Best, Sebastian

fazeliniah commented 2 years ago

Thanks for your quick response. I think it makes sense to keep the imputation. The bar plots for the protein groups is a wonderful way to show the summary. But maybe having additional tab that shows presence/absence for detected proteins across groups can add more useful information. something like this for proteins before imputation: Example_venn

tbaccata commented 2 years ago

I agree, this would be a good QC feature. I will implement this.

Best, Sebastian

fazeliniah commented 2 years ago

Thanks! Look forward to the new feature.

tbaccata commented 2 years ago

Hello,

To keep this feature general for many samples, and because set comparisons become very difficult to visualize for many sets, I decided to opt for a heatmap.

You have 3 options to compare the overlap between samples in this heatmap (if you click on the wrench icons):

overlap_heatmap

In addition, you can inspect the values of these metrics in a data table. These features are available in the "Protein Overlap" tab next to the Protein groups barplots.

Best, Sebastian