Very large set of transcripts grouped in one splice graph

[KristinaGagalova]

Hi,

We are testing Chopstitch with known transcripts.

We are using Homo_sapiens.GRCh38.ensembl_94.cdna with transcripts > 200 bp and redundancy removed at 98% (cd-hit-est). The reference genome is NA12878 - reads.
This is the output that we have after FindSubcomponents.py
geneMap.tsv.gz

There is a huge cluster of 7429 transcripts grouped together. I attach the fasta file with those sequences:
largeCluster.fasta.gz

This looks a bit odd. Usually with every run of Chopstitch there is a very big cluster that groups together transcripts with very dissimilar sequences
Any clues where this can come from?
Thanks,

[hamzakhanvit]

MakeSplicegraph.py groups together input putative exons that have similar suffixes and prefixes (30 bases). Even if one exon from a transcript has a similar prefix or suffix to another exon from another transcript, they would be grouped together. This might result in erroneous groups and is something that I want to improve upon. I am planning to use cdhit to group together putative exons and then parse its result to generate a splice graph.