Can you swap reference genome releases ?

Question

[jonas.andersson]

I'm aware that you have to stick to the same reference genome during the whole analysis. I.e., if the analysis started with hg19, you have to stick to that to not introduce inconsistencies.

However, is it the same with releases for the same assembly?

If you have mapped the reads to e.g grch38.p13 release 32, and a new update is released e.g, grch38.p13 release 43, which is the newest. Can you change to that without having to remap the reads to get new bam files? What if you change from grch38.p12 to grch38.p13?

As far as I have understood, the gene coordinates are not changed in these releases. Can someone please clarify what is correct here, Im so confused 🙃

Thanks in advance!

[GenoMax]

Patch releases do not change main chromosome coordinates. So you don't need to worry about that for normal analysis as in you will not need to remap the reads. If you were working with alt regions, haplotypes then those would be added/amended in each patch.

That said it is best to stay with a consistent reference (patch and all) for one set of analyses.

[jonas.andersson]

Great, thx for you straight forward answer!

Answer 1

[Istvan Albert]

You can transform coordinates from one assembly to another, but in general you cannot transfer alignments from one assembly to another.

it goes back to that difference between mapping and aligning.

If one only needs to work with coordinates, those can be transferred with a transformation: coordinate X becomes coordinate X + N , but the alignments contain much more information and are not amenable to a simple linear transformation.

Long story short, you cannot transform a BAM file onto a new assembly.

[benformatics]

Interesting statement about the semantics of mapping vs. aligning. Can you elaborate (just out of curiosity)? Are you saying an alignment is mapping with extra meta-data?

[Istvan Albert]

I first saw the distinction between mapping and alignment in some slides by Heng Li, from .... looks like 2011 ... ahem ... that stuck in my mind ever since

https://www.broadinstitute.org/files/shared/mpg/nextgen2011/nextgen2011_li.pdf

the interesting thing is that with mapping and alignment each could be correct or incorrect, regardless of the state of the other

[jonas.andersson]

very interesting! thx for sharing! btw, I just realised you are the author of the biostar handbook I recently bought! thx for a great book! However, I have found some typos and such, where can I report it to be corrected for future releases? best Jonas

[Istvan Albert]

typos are plentiful, don't worry about them :-) there are literally thousands of pages of content, with spotty proofreading because that is a job on its own - and I really dislike proofreading

[jonas.andersson]

But I'm a person that reads a book from first to last page and also takes notes on every error I come upon (part of my ocd i guess😆).Then you might as well make use of it😄

[jonas.andersson]

Thx for you answer!

Answer 2

[ATpoint]

Are you referring to Ensembl/GENCODE releases? The actual reference genome sequence is defined by the assembly (GRCh38) and its current patch (here p13). Patches correct sequence errors etc but do not change coordinates.

In any case, if you're referring to aforementioned releases then the genome is the same. What changes slightly in these releases are gene annotations but not the genome. GENCODE 32 to 43 all use the same genome version which is GRCh38.p13.

https://www.gencodegenes.org/human/releases.html

[jonas.andersson]

thx for you answer! lets say I have used GRCh38.p13 release 43 until now and GRCh38.p13 release 44 i released tomorrow. Can I easily switch to the latest release without anything going wrong? But what would be the reason to do it? is it worth to always stick to the latest release?

[ATpoint]

You ask me why you consider switching? You have to know. Basic rule is to be consistent within project. Use the same for everything.

[jonas.andersson]

No that was not what I meant 😄 I just assume that it allways best to use the latest release, but maybe that is not that important? Sorry if I sound obtuse, I'm still a rookie 😊