icbi-lab / NeoFuse

NeoFuse is a user-friendly pipeline for the prediction of fusion neoantigens from tumor RNA-seq data.
GNU General Public License v3.0
17 stars 9 forks source link

The results were very different from GRCH37 and GRCH38. #22

Closed renyuan001 closed 5 months ago

renyuan001 commented 5 months ago

I use 10 samples and analysed both by GRCH37 and GRCH38. I conpared the ***filtered.tsv and unfiltered.tsv between GRCH37 and GRCH38. The fusion genes and locations were all different. No same result were found and it seems like the result referenced by GRCH38 was much fewer than the result referenced by GRCH37. Please tell me the reason, thank you! $ NeoFuse -R -o GRCH37 -n 20 -V GRCH37 --docker $ NeoFuse -R -o GRCH38 -n 20 -V GRCH38 --docker These orders were used to generate reference.

abyssum commented 5 months ago

Hello @renyuan001,

First of all, thank you for using NeoFuse.

To briefly answer your question, since GRCh38 (hg38) is and update of GRCh37 (hg19), there are several thousand individual bases updated in GRCh38, many of which corrected errors in coding sequences. Since this is beyond the scope of NeoFuse, I will not elaborate on the differences between the two versions (there is plenty of information available, this publication for example).

Although, I haven't tried to compare the results from the two versions myself, I would expect that you would find major differences between GRCh37 and GRCh38 as you mentioned (especially when it comes to genomic positions). As a quick example, consider BRCA2:

Following the previous example, let's say that the majority of the sequenced reads of your sample fall under the genomic position chr13: 32,315,086-32,400,268. Then depending on the annotation you would be using, BRCA2 may be called (using GRCh38) or not (when using GRCh37).

Given the fact that in order for Arriba to call a fusion the reads should span the junction point (for more info please read the arriba documentation regarding split reads, discordant mates and coverage), it is expected that since GRCh38 is more robust it will - probably - yield less results (meaning that GRCh37 might yield more results which are potentially false positives).

I hope this helps.

renyuan001 commented 5 months ago

Thank you for your kind reply and I will use GRCh38 to analysis my samples. It is a good tool for screening neoantigen from genefusion.

abyssum commented 5 months ago

I am glad I could help.

Once again thank you for supporting NeoFuse and please let me know if you encounter any issues.

As a side note, we will be updating NeoFuse to a newer version in the following months. Some of the new features include:

You can "watch" the current repo version to get notification for future updates.