Cannot simulate transcriptomic reads with intron retention

[SimonHegele]

Hello,

I was trying to simulate transcriptomic nanopore reads.

First I created a model for the mouse using the following command:
read_analysis.py transcriptome -i reads/long/SRR11517552.fastq -rg reference/genome.fna -rt reference/transcriptome.fna -annot reference/annotation.gff -o reads/long/SRR11517552_nanosim/model -t 128
Organism: mus musculus
Genome: GRCm39
Reads: SRR11517552 (from NCBI-SRA, single-cell transcriptomic, blastocyst)

I then tried to simulate reads using the model and an expression profile I created with Kallisto from Illumina short reads I found in the NCBI-SRA ( PRJNA375882) using the following command:
simulator.py transcriptome -rt reference/transcriptome2.fna -rg reference/genome.fna -e reads/short/abundance_nanosim.tsv -c reads/long/SRR11517552_nanosim/model -o reads/long/nanosim -t 128

NanoSim reads all input files but gets stuck at the simulation step. If I add the --no_model_ir parameter reads are simulated.
The issue may originate from the modelling step as the training files look suspicious.

model_IR_info.txt
model_IR_markov_model.txt

Best,
simon

[lcoombe]

Hi Simon,

Would you be able to provide me with some more information? Namely, the version of NanoSim used and the full log from your read_analysis.py step? Indeed, it seems like there was an issue in the modelling of intron retention, but it would be helpful to see that extra info.

Thank you for your interest in NanoSim!
Lauren

[SimonHegele]

Thanks for the quick response!

The version is 3.2.2, the log-file is attached.

nanosim.log

[lcoombe]

Thanks Simon.

Do your transcript names start with ENST? Or do they use a non-ENSEMBL convention?

It could be related to this issue: #156

[SimonHegele]

Transcript-IDs look like this: ID=rna-NM_001011874.1 (in the annotation) or >NM_001011874.1 (in the transcriptome-fasta).

For the simulation I had to remove the underscore in the transcript-IDs in the transcriptome-fasta (see Issue #237).

[lcoombe]

Hi Simon,

I think the naming is the issue for the intron retention step here.
I believe there are two issues at play:

1. The IDs in the annotation have the rna- prefix, so don't match exactly with the transcripts fasta file, as expected
2. I looked at the model_intron_retention.py script, and it looks like it expects the transcripts to start with the Ensembl convention "ENST". There isn't really an algorithmic reason for why this should be needed as far as I know, but it is in the code:

   NanoSim/src/model_intron_retention.py

   Line 94 in 1ae603e

   if tname.startswith('ENST'):

If possible, I would try ensuring that the IDs totally match in the two files, and add the ENST prefix, and see if that gives you proper files after the training.