Psi Not Calculated for Event Error: Similar to Issue #201

[stevenjbark]

Hello SUPPA Members,

I have a similar problem as was described for Issue #201 and wanted to get clarification of this problem with an alternative application. I have generated data in single-cell RNASeq experiments and was able to export the data into a counts format compatible with SUPPA 2.4. I used both the FASTA database and the GTF file for the same versions, so they should be compatible. After generating the .ioe files and the expression file, I was able to use the psiPerEvent function in suppa.py without errors. However, every single entry was NA, as was described for Issue #201. Recall that single-cell experiments can measure many different transcripts and annotate these to particular cell types, but typically do not have measurements for every single transcript in each cell type or experiment.

In discussing this issue, it was noted that every transcript in the last column of the .ioe file must be present in the expression file. Is this correct, even if one or more of the transcripts in the .ioe file are actually in the expression file, but not all? There may be valid biological reasoning for this, but that would also limit the application of SUPPA for alternative splicing analysis in single-cell RNASeq experiments.

Regards,
Steven J. Bark, Ph.D.
Bioinformatics Analyst
Baylor College of Medicine

[EduEyras]

Hi There is no problem applying SUPPA to single cell data. We've done this before. It is just like any other analysis of single-cell or bulk sequencing data, if you don't have reads for a transcript, you could assume zero expression, or some background distribution, and let the method use that. The key here is that the generation of the .ioe uses all transcripts in the GTF before looking at their expression. But the expression file is produced independently, so what's in the expression file may depend on what you used to generate it. You could add all missing transcripts with zero value. Issue #201 could be due to hidden symbols. I hope this helps Eduardo

…

On Wed, 16 Apr 2025 at 05:49, Steven J Bark ***@***.***> wrote: Hello SUPPA Members, I have a similar problem as was described for Issue #201 <#201> and wanted to get clarification of this problem with an alternative application. I have generated data in single-cell RNASeq experiments and was able to export the data into a counts format compatible with SUPPA 2.4. I used both the FASTA database and the GTF file for the same versions, so they should be compatible. After generating the .ioe files and the expression file, I was able to use the psiPerEvent function in suppa.py without errors. However, every single entry was NA, as was described for Issue #201 <#201>. Recall that single-cell experiments can measure many different transcripts and annotate these to particular cell types, but typically do not have measurements for every single transcript in each cell type or experiment. In discussing this issue, it was noted that every transcript in the last column of the .ioe file must be present in the expression file. Is this correct, even if one or more of the transcripts in the .ioe file are actually in the expression file, but not all? There may be valid biological reasoning for this, but that would also limit the application of SUPPA for alternative splicing analysis in single-cell RNASeq experiments. Regards, Steven J. Bark, Ph.D. Bioinformatics Analyst Baylor College of Medicine — Reply to this email directly, view it on GitHub <#211>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADCZKB25ICBG6KY3XJRFPTT2ZVPEBAVCNFSM6AAAAAB3GPZEHSVHI2DSMVQWIX3LMV43ASLTON2WKOZSHE4TONBVGA4DQMA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> *stevenjbark* created an issue (comprna/SUPPA#211) <#211> Hello SUPPA Members, I have a similar problem as was described for Issue #201 <#201> and wanted to get clarification of this problem with an alternative application. I have generated data in single-cell RNASeq experiments and was able to export the data into a counts format compatible with SUPPA 2.4. I used both the FASTA database and the GTF file for the same versions, so they should be compatible. After generating the .ioe files and the expression file, I was able to use the psiPerEvent function in suppa.py without errors. However, every single entry was NA, as was described for Issue #201 <#201>. Recall that single-cell experiments can measure many different transcripts and annotate these to particular cell types, but typically do not have measurements for every single transcript in each cell type or experiment. In discussing this issue, it was noted that every transcript in the last column of the .ioe file must be present in the expression file. Is this correct, even if one or more of the transcripts in the .ioe file are actually in the expression file, but not all? There may be valid biological reasoning for this, but that would also limit the application of SUPPA for alternative splicing analysis in single-cell RNASeq experiments. Regards, Steven J. Bark, Ph.D. Bioinformatics Analyst Baylor College of Medicine — Reply to this email directly, view it on GitHub <#211>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADCZKB25ICBG6KY3XJRFPTT2ZVPEBAVCNFSM6AAAAAB3GPZEHSVHI2DSMVQWIX3LMV43ASLTON2WKOZSHE4TONBVGA4DQMA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[naazkhan]

Hi, I am getting the same error, and so far, no luck. All entries are NA, with PSI not calculated. The TPM files without any expression data are marked 0; still, the value of PSI is NA. Any help is highly appreciated.

Naazneen

[EduEyras]

Is this for single cell data?

If the transcript expression is very sparse, there will be many zeroes, which may result in NA's for the PSI.

Can you do any imputation to rescue TPM expression values?

E.

[naazkhan]

This is for Bulk seq, not single cell. I did the imputation of TPM values, but still no result.

[EduEyras]

The PSI can only be calculated if at least one of the transcripts in the last column of the .eoi file has a non-zero value, otherwise it is not defined.

Did you confirm that there are events with all transcripts having non-zero value? Those should not be NA.

Do you get any error during the run?

E.