search

richelbilderbeek / bbbq_article

The Bianchi Bilderbeek Bogaart Question answered
GNU General Public License v3.0
0 stars 0 forks source link

[bbbq_article_issue_157] Do stats on TMH presentation of elution studies #230

Closed richelbilderbeek closed 3 years ago

richelbilderbeek commented 3 years ago

Needed for #228, where is now wirtten:

The authors do present an example of TMH that are known to be presented by MHC-I and MHC-II, from two peptide elution studies. These data indeed appear to show that TMH are found among the eluted peptides (“appears” because the figure is missing). However, the authors do not provide any analysis of whether the frequency of such epitopes is more or less than would be expected by chance, as they do in detail for the predicted epitopes, just the presence of TMH-derived epitopes in the peptidome is noted.

The reviewer is right that we do not show any statistics here; we were simply already convinced that 1.3% (for MHC-I) and 3.9% (for MHC-II) of the in vivo presented epitopes to be of (predicted) TMH origin, as the expected percentages would be zero percent. We agree with the reviewer and we did [some statistical tests]

richelbilderbeek commented 3 years ago

I guess we need to prove the percentage is not zero.

richelbilderbeek commented 3 years ago

bbbq_article_issue_157 it is:

results

richelbilderbeek commented 3 years ago

Screenshot from 2021-10-18 08-05-58

richelbilderbeek commented 3 years ago

Replied:

The reviewer is right that we do not show any statistics here and we are happy this omission was spotted. However, unlike most omitted statistics tests, this would be an exception when adding a statistical test would not make sense.

To clarify, we considered adding the text below to describe the statistical test:

We used a binomial statistical test to determine if more TMH-derived epitopes are presented than expected, with the number of epitopes being the number of trails, the number of TMH-derived epitopes being the number of successful trials, for a estimated chance of success of zero. This resulted in highly significant values, indicating there were seignificantly more TMH-derived epitopes presented than expected.

We still feel it would be better to leave out this trivial statistics test, as it sets readers on the wrong foot: any binomial test with a chance of a successful trial, p, of zero (in R code: binom.test(x = 109, n = 7897, p = 0.0)) always results in a p value of zero.

richelbilderbeek commented 3 years ago

Fixed with 3cbbb1cca578341fcce51cd7218d711dcede9048