galig
galig copied to clipboard
AlgoLab
ALK ATI event
Hello everyone, I have been using ASGAL for some time now and I'm very content with the obtained results, congrats on the implementation. Lately I have been working with samples that present the ALK ATI isoform. ASGAL hasn't been successful at calling this event. After running the program with a few samples (all having this alteration) I have started to think that ASGAL may not be designed to identify this type of event (based on my interpretation of the documentation), but before I jump into that conclusion I would like to know your thought on this. This image provides a nice description of the ALK ATI event. Let me know if I should provide additional information. Thanks in advance!
Hi, thanks!
Just to be sure: in the event reads map to intron 19 as in Figure 1, right?
If that's the case, ASGAL struggles a bit in identifying such events. When the portion of the intron covered by reads is small enough (and a relatively long portion of the read can be placed on the exons), ASGAL should be able to detect the event. But when that portion is big (or it's the entire intron), ASGAL cannot identify it. Indeed, ASGAL starts by aligning reads to the exons and then detects events by analyzing the alignments. If a read must be placed entirely on an intron or it aligns more on the intron (and not sufficiently long anchors can be placed on the exons), ASGAL do not map it (and the event is lost).
Let me know if this is clear enough.
Thanks for the quick reply : ) The sample alignments shows reads mapping to exon 20 and intron 19. Some of them even partially mapping both features. The reason I started to suspect this event might not be detected by ASGAL is due to the fact that none of the reads involved in the event are spliced, meaning that there is no partial read alignment with exon 19, in fact no reads are aligned to this exon. For this novel isoform exon 20 is actually the first exon of the transcript. Is it possible that this could also be penalizing the calling procedure? Given that you have already identified this limitation is there any plan on covering this scenario and make ASGAL capable of calling such events? In case you are not planning on adding this features, would you suggest a different approach/program to complement ASGAL and identify this type of alteration? Thanks again!
If I understand correctly you have reads that are not spliced and covers exon 20 and intron 19. If you want to identify this isoform in a sample or a novel isoform (in a general case) you should have it in the splicing graph and then run ASGAL with the novel isoform as an annotated one. Luca can tell you how you ca do it for the moment. Yes, the idea is to extend the tool to include intron portions. Anyway the general idea should be that if you conjecture a given isoform that is not annotated but present in the sample, then you should give it to ASGAL as input in the annotation (used to build the splicing graph). Then ASGAL should be able to detect it, not as novel but as an annotated one in the sample
Hello again, thank you both for your comments. @bonizzoni that is correct, I have full length reads overlapping simultaneously exon 20 and intron 19. This is observed only for those samples with a confirmed case of an ALK ATI event. I can give it a try and perform the steps you suggest, but of course, any additional guidance is always welcomed. Also, very happy to see you are planning on adding more features to ASGAL, looking forward to that! Bests!
Hi @ldenti and @bonizzoni. I am working with @gremame on this issue. We tried to implement what you recommended, but do not understand well how it fits in this situation.
The ALK ATI event is already annotated. I attach below an IGV screenshot with the reference hg38, including genome and annotated features, and the reads aligned by the SpliceAwareAligner of ASGAL.
As you can see and @gremame explained, there are reads overlapping intron and exon as we were expecting. However, we do not understand what you mean by:
If you want to identify this isoform in a sample or a novel isoform (in a general case) you should have it in the splicing graph and then run ASGAL with the novel isoform as an annotated one.
Anyway the general idea should be that if you conjecture a given isoform that is not annotated but present in the sample, then you should give it to ASGAL as input in the annotation (used to build the splicing graph). Then ASGAL should be able to detect it, not as novel but as an annotated one in the sample
This isoform is present in the sample and annotation but is not being detected. Maybe ASGAL cannot deal with this type of event?
We considered using a more specific Alternative Translational Initiation (ATI) tool, such as preTIS or TITER but are not very convinced and prefer to continue using ASGAL if possible.
I see your point. From the IGV screenshot, I can confirm that unfortunately ASGAL cannot detect that event. Indeed, you don't have any spliced alignments and all the events ASGAL can detect comes from spliced alignments (it doesn't use coverage information). Moreover, even if you "merge" the two exons and the intron in a single exon, you will get the same alignments you are getting: the part of the exon you are interested in is already annotated. Finally, I would classify that event as a transcriptional event (and asgal was designed to detect alternative splicing events).
I think we could be able to modify asgal to detect that event but it will take some time (in case, would it be possible to you to share the data?).
Just a curiosity: do you have any clue why there is no spliced alignments overlapping that introns but you have reads covering the two exons? I don't see any transcript starting with the "last" exon I see in the screenshot. Maybe ASGAL is missing some alignments there?
Thank you for your response @ldenti. We are currently asking for permission from the owner of the data to share the data with you. How shall we send it to you?
Regarding your enquiry, we believe there is no read overlapping both intron and "last" exon because in this event the intron marks the beginning of the new isoform (Alternative Translational Initiation, ATI). Therefore in this isoform there would be no reads mapping to the "last" exon. In this sample, those two reads on the exon are likely "contamination" from the other isoform, although it seems that the ATI isoform is the predominant isoform in this sample.
Ok, got it! Thanks!
Actually I just need the annotation and the reads falling on this gene (fastq used with asgal or the output sam). They should be small: you can send me an email with a drive/dropbox/... link.