Dear developers,
At my lab, we are currently developing a workflow to analyze target-enrichment data for phylogenomics of polyploids. Saute is a crucial step in the pipeline, and we are quite happy with its performance.
A key aspect of how we are using Saute is that we feed it with a redundant set of seeds to maximize sensitivity. The reason is that the homeologs we expect in the data can be very similar in the exons, but vary much more in the introns or promoters. We have our way to generate this "enriched" set of seeds, from which a subset is probably the real set of variants for the given sample, and we trust Saute to knock out the redundant/uninformative ones by incorporating the read and pair information.
At the start, we were using the --selected_variants output, since we understood that it yields a subset of the best supported variants reported in --all_variants output. However, recently we noticed that in some instances, quite often in real data, --selected_variants produced many more sequences than --all_variants.
The above observation rocked our priors, since we thought that --selected_variants would always be a subset of --all_variants. I also noticed that these big differences are driven by a few seeds for which Saute discovers an insane number of variants. These variants are limited by the --max_variants parameter (we use 10) for the --all_variants output, but I suspect the parameter does not limit the --selected_variants output (Do you think this might be the case?).
So we are currently in the process of evaluating which output is more convenient for our application, and we have come to the point where we have to admit we do not get all the nuances and differences between the two outputs to make a proper decision.
So, can I ask you to please elaborate on the nuances to generate both outputs? particularly considering the way we are using Saute? That is, using a reference set with "redundancy" (where probably some of the variants to be found are already present).
Please don't feel shy to go as deep and technical as you need. We have read the paper multiple times and have been playing with Saute for months now. We believe we have a good grasp on it, but there is simply not much information about the --selected_variants part.
If you have any questions or would like me to provide some concrete examples (numbers, data, etc.), please let me know.
Thank you in advance for your help,
Sivico
Dear developers,
At my lab, we are currently developing a workflow to analyze target-enrichment data for phylogenomics of polyploids. Saute is a crucial step in the pipeline, and we are quite happy with its performance.
A key aspect of how we are using Saute is that we feed it with a redundant set of seeds to maximize sensitivity. The reason is that the homeologs we expect in the data can be very similar in the exons, but vary much more in the introns or promoters. We have our way to generate this "enriched" set of seeds, from which a subset is probably the real set of variants for the given sample, and we trust Saute to knock out the redundant/uninformative ones by incorporating the read and pair information.
At the start, we were using the
--selected_variantsoutput, since we understood that it yields a subset of the best supported variants reported in--all_variantsoutput. However, recently we noticed that in some instances, quite often in real data,--selected_variantsproduced many more sequences than--all_variants.The above observation rocked our priors, since we thought that
--selected_variantswould always be a subset of--all_variants. I also noticed that these big differences are driven by a few seeds for which Saute discovers an insane number of variants. These variants are limited by the--max_variantsparameter (we use 10) for the--all_variantsoutput, but I suspect the parameter does not limit the--selected_variantsoutput (Do you think this might be the case?).So we are currently in the process of evaluating which output is more convenient for our application, and we have come to the point where we have to admit we do not get all the nuances and differences between the two outputs to make a proper decision.
So, can I ask you to please elaborate on the nuances to generate both outputs? particularly considering the way we are using Saute? That is, using a reference set with "redundancy" (where probably some of the variants to be found are already present).
Please don't feel shy to go as deep and technical as you need. We have read the paper multiple times and have been playing with Saute for months now. We believe we have a good grasp on it, but there is simply not much information about the
--selected_variantspart.If you have any questions or would like me to provide some concrete examples (numbers, data, etc.), please let me know.
Thank you in advance for your help,
Sivico