Some fusions detected in subsampled reads are missing when using full dataset in ctat-LR-fusion

[CyberGypsy6324]

Hi,

I am using ctat_LR_fusion for gene fusion analysis. For the same sample, I ran the analysis three times using 1/3 of the reads, 2/3 of the reads, and the full dataset. When comparing the results (ctat-LR-fusion.fusion_predictions.abridged.tsv), I noticed something surprising: some fusions detected in the 1/3 or 2/3 reads runs are not detected in the full dataset run. This pattern also appears in ctat-LR-fusion.fusion_predictions.preliminary.abridged.tsv.

Could you please show me what caused it?

Thank you!

[brianjohnhaas]

Hi, It's likely due to this parameter used for filtering: # --min_FFPM <float> :min fusion expression (default: 0.1 FFPM - meaning at least 1 fusion read per 10 M total reads ) Try setting it to zero or a very small number to see if that helps. best, Brian

…

On Tue, Aug 26, 2025 at 7:35 PM Li Gan ***@***.***> wrote: *CyberGypsy6324* created an issue (TrinityCTAT/CTAT-LR-fusion#15) <#15> Hi, I am using ctat_LR_fusion for gene fusion analysis. For the same sample, I ran the analysis three times using 1/3 of the reads, 2/3 of the reads, and the full dataset. When comparing the results (ctat-LR-fusion.fusion_predictions.abridged.tsv), I noticed something surprising: some fusions detected in the 1/3 or 2/3 reads runs are not detected in the full dataset run. This pattern also appears in ctat-LR-fusion.fusion_predictions.preliminary.abridged.tsv. Could you please show me what caused it? Thank you! plot_zoom_png.png (view on web) <https://github.com/user-attachments/assets/64c49d52-b760-46ea-bfdf-9aaebdd00dd9> — Reply to this email directly, view it on GitHub <#15>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABZRKX3JDUDAOCXCH3SZ6LT3PTVNJAVCNFSM6AAAAACE4RDSS6VHI2DSMVQWIX3LMV43ASLTON2WKOZTGM2TONBUGAZTONQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- -- Brian J. Haas The Broad Institute http://broadinstitute.org/~bhaas <http://broad.mit.edu/~bhaas>

[CyberGypsy6324]

I'm sorry I forgot to mention that the parameter --min_FFPM 0 was already set when running. Here is my script:

ctat-LR-fusion \
-T data.fastq \
--genome_lib_dir GRCh38_gencode_v44_CTAT_lib_Oct292023.plug-n-play/ \
--CPU 24 \
--vis \
--min_FFPM 0 \
--examine_coding_effect \
--no_annot_filter

Thank you!

[brianjohnhaas]

I’m not sure then. Another thing you could try is to feed it aligned bam files where you subset aligned bans instead of fastqs. Then if the issue persists you could maybe share the bans with me for troubleshooting further.

…

-- Brian J. Haas The Broad Institute http://broadinstitute.org/~bhaas <http://broad.mit.edu/~bhaas>

On Wed, Aug 27, 2025 at 3:53 PM Li Gan ***@***.***> wrote: *CyberGypsy6324* left a comment (TrinityCTAT/CTAT-LR-fusion#15) <#15 (comment)> I'm sorry I forgot to mention that the parameter --min_FFPM 0 was already set when running. Here is my script: ctat-LR-fusion \ -T /scratch/yzhan937/epi2me/wf_transcriptomes/BC15T_3/run_1/output/BC15T_3_pychopper_output/BC15T_3_full_length_reads.fastq \ --genome_lib_dir /scratch/yzhan937/epi2me/ctat_lr_fusion/ctat_genome_lib_build_dir/GRCh38_gencode_v44_CTAT_lib_Oct292023.plug-n-play/ \ --CPU 24 \ --vis \ --min_FFPM 0 \ --examine_coding_effect \ --no_annot_filter Thank you! — Reply to this email directly, view it on GitHub <#15 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABZRKX2ZOHFYL33W6TEZM533PYEFJAVCNFSM6AAAAACE4RDSS6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTEMRZGU2TKNRVGA> . You are receiving this because you commented.Message ID: ***@***.***>

[brianjohnhaas]

I'll run some of my own data through in a similar fashion to see if the same happens and will explore it further. More later.

[brianjohnhaas]

Hi, I ran a similar test using the SKBR3 pacbio data we published on like so: ctat-LR-fusion --CPU 10 -T DepMap_AS010_v1_SKBR3.refine.bam.fastq.gz --no_annot_filter --min_FFPM 0 -o SKBR3_all ctat-LR-fusion --CPU 10 -T DepMap_AS010_v1_SKBR3.refine.bam.fastq.gz.tab.two_thirds.fastq --no_annot_filter --min_FFPM 0 -o SKBR3_twothirds ctat-LR-fusion --CPU 10 -T DepMap_AS010_v1_SKBR3.refine.bam.fastq.gz.tab.one_third.fastq --no_annot_filter --min_FFPM 0 -o SKBR3_onethird with results of fusion call counts: full.fusions 466 full.fusions,twothirds.fusions 185 full.fusions,twothirds.fusions,onethirds.fusions 177 and so didn't find any that were unique to the smaller one-third or two-third runs. All are proper subsets. I do see that it looks like you're using ONT reads instead of pacbio. The annot filter step does have some logic around weeding out fold-back types of fusions that show up as artifacts from fold-back reads in ONT and maybe this is impacting your findings, but I'm not entirely sure. hope this helps, Brian

[CyberGypsy6324]

Hi Brian,

Yes my data is ONT nanopore cDNA. I will try to look into the intermediate files and adjust the parameters.