starfish dereplicate output help

[drew-olson-tmru]

Hi,

I am following the tutorial with my own dataset of Fusarium genome data (six F. fujikuroi and six F. proliferatum) and I'm not entirely sure what the log from dereplicate means:

[Tue Sep 30 13:59:21 2025] reading in data..
[Tue Sep 30 13:59:21 2025] parsing 12 gff files for all mRNA features and ignoring everything else..
[Tue Sep 30 13:59:22 2025] parsing 3 flanking OGs within 600000bp of the up and downstream boundaries of 10 elements..
[Tue Sep 30 13:59:22 2025] failed to retrieve the flanking haplotypes of 10 elements with fewer than 2 OGs on at least one flank
[Tue Sep 30 13:59:22 2025] grouping the flanks of the remaining 0 elements into homologous region groups, allowing for a maximum of 1 OG mismatches..
[Tue Sep 30 13:59:22 2025] grouped element flanks into 0 homologous region groups, where 0 region groups have >1 element and 0 have 1
[Tue Sep 30 13:59:22 2025] searching for haplotypes missing element insertions but with homology to region groups..
[Tue Sep 30 13:59:22 2025] found 0 empty haplotypes distributed across 0 homologous region groups..
[Tue Sep 30 13:59:22 2025] searching for fragmented haplotypes missing element insertions but with homology to known region groups..
[Tue Sep 30 13:59:22 2025] found 0 fragmented haplotypes across 0 homologous region groups..
[Tue Sep 30 13:59:22 2025] merging regions where >=50% of empty or frag haplotypes overlap by >=25% of their length..
[Tue Sep 30 13:59:22 2025] after merging: grouped element flanks into 0 homologous region groups, where 0 region groups have >1 element and 0 have 1
[Tue Sep 30 13:59:22 2025] after merging: found 0 empty haplotypes distributed across 0 homologous region groups..
[Tue Sep 30 13:59:22 2025] after merging: found 0 fragmented haplotypes across 0 homologous region groups..
[Tue Sep 30 13:59:22 2025] cross-referencing the coordinates of 94 element insertion sites with empty and fragmented haplotypes..
[Tue Sep 30 13:59:22 2025] validated 0 insertion sites found inside empty and fragmented haplotypes
[Tue Sep 30 13:59:22 2025] cross-referencing the coordinates of 25 tyrs not associated with elements with the coordinates of fragmented haplotypes..
[Tue Sep 30 13:59:22 2025] found 0 tyrs inside fragmented haplotypes
[Tue Sep 30 13:59:22 2025] choosing one representative element per element family at each of the 0 homologous region groups..
[Tue Sep 30 13:59:22 2025] found 0 regions with at least 1 cross-referenced element-insertion site pair
[Tue Sep 30 13:59:22 2025] found 0 regions with 1 element and >=1 empty/fragmented haplotype
[Tue Sep 30 13:59:22 2025] found 0 regions with >1 element and >=1 empty/fragmented haplotype
[Tue Sep 30 13:59:22 2025] found 0 regions with 1 element and 0 empty/fragmented haplotypes
[Tue Sep 30 13:59:22 2025] found 0 regions with >1 element and 0 empty/fragmented haplotypes
[Tue Sep 30 13:59:22 2025] calculating median Jaccard similarity between all haplotypes and all regions..
[Tue Sep 30 13:59:22 2025] done

I cannot continue following along the tutorial steps after this since there is nothing in the output files. Can someone help explain this to me?

Thank you!

[egluckthaler]

hi! it looks like the program could not retrieve the flanking regions for any of the 10 predicted elements that you had found "failed to retrieve the flanking haplotypes of 10 elements with fewer than 2 OGs on at least one flank". The program should really exit at this point because their is nothing left to do if no flanking regions are retrieved. In cases like this, it usually means that the quality and contiguity of the assemblies are very poor in the regions in which the starships are located. I wonder if generally your assemblies have low N50s, I suspect that using higher quality assemblies would fix the issue.

[drew-olson-tmru]

Hi, thank you for getting back to me. So, my assembly contigs/scaffolds range between 12 (chromosome level) - 18 and the N50's > 4Mb for each. This is probably a me problem regarding gff3 and annotation processing...

My genomes are completely annotated with Funannotate (all have their respective NCBI locus prefix tags for unique names). I was having an issue with the gff3 files and eggnog annotations from the Funannotate output at this same step too. Originally, this part of the pipeline would not run because it would tell me that I did not have any mRNA data in the gff file. I thought this was because Funannotate provides the "gene_name-T1" or whatever number of isoform for the mRNA data and it looked that was causing the issue. In the most recent run, I extracted just the mRNA data, kept a single isoform, and removed the "-T1" from the gff file in effort to make it look like your example gff file, and I removed the same things from the eggnog annotations, but I'm guessing that probably caused this issue.

Is there some additional processing that I need to do with the Funannotate data prior to running this pipeline, or do you have some recommended steps when working with Funannotate data? I am no computational biologist, so odds are that I'm breaking something in my data.

P.S. I apologize for the lack of actual log files and snap shots of the data used. I am with the USDA and cannot access that data for the time being... I have this program up and running on my associated university's HPC, so I will be able to work on this again soon.

[egluckthaler]

Thanks for the additional info, that's very helpful. when starfish parses GFF files, it only looks at specific features (default: mRNA, but you can change this in most scripts with the --targetFeat switch). It then parses the names of those features from a specific field (default: 'Name=', but you can change this with the --nameField switch). Are you still getting errors saying that there are no 'mRNA' features to parse in the file? I suspect that there is an issue with the type of feature that you are trying to parse and the field where the feature name is located. When you have access to your data again, it would be helpful for me to see it to give more specific advice. Note that if you try editing the names of the features halfway through the pipeline (e.g. removing the -T1) then that can cause all sorts of errors because the names of the features in the BED files that starfish produces will no longer match the names in the GFF