kevinmhadi
commented
4 days ago Threshold Specification: @kevinmhadi will provide the specification of the QC metrics object's attributes and their respective thresholds (in this issue thread) to indicate bad samples. For now these will be flat comparisons, later we can decide on more sophisticated criteria for flagging.
These values will be subject to change, as many final sequencing specs aren't even decided on yet. But I could see flagging might have 2 levels where if any of the following are seen, these are either serious signs of bad sequencing ("danger") or less serious, but still concerning issues over data quality ("warning").
I think these are the stats where any of these issues should flag one of these two levels of issues (doesn't have to actually be named "danger" or "warning").
The below is just an example mapping of what the category is that would raise one of the two levels of severity and what a potential message should show in a tooltip, but I am imagining the UI element would end up on one of three options ("green light", "red light"/Danger, "yellow light"/Warning). If there's a mixture, the UI element should defer to the most serious level.
Danger
{
"Tumor Coverage < 80X": "Tumor bam has low coverage: 23X",
"Normal Coverage < 20X": "Normal bam has low coverage: 10X",
"Mapped Reads / Total Reads <99%": "High percentage of reads not mapping to hg19: X%",
"Mapped HQ (>=Q20 Reads) / Total Reads <80%": "High percentage of reads not mapping to hg19: X%",
"Genome Coverage at least 30X (PCT_30X, Picard) is less than 95%": "Insufficient coverage of genome at 30X: X%",
"AMBER Contamination > 1%": "Potential sample contamination detected!"
}Warning
{
"Median Insert Size <= 300": "Short median insert sizes detected: sensitivity to structural variants may be reduced.",
"Optical/PCR Dups >= 20%": "High duplication rate X%",
"JaBbA: Number of Loose Ends >= 100": "Sample potentially hypersegmented",
"JaBbA: Converged == FALSE": "Graph model did not converge, PGV CN solution must be interpreted with caution"
}
shihabdider
xanthopoulakis
This feature involves adding a QC tooltip to the case metadata header.
Feature Details:
Tooltip Design:
The tooltip should be consistent with existing tooltips in the header:
Icon Indication:
An icon should be associated with the tooltip to indicate the quality status of the case. This icon will reflect whether the case meets predefined thresholds for QC metrics. @xanthopoulakis feel free to use defaults provided by ant.design.
Data Source:
QC metrics will be sourced from the
metadata.jsonfile under the attributebam_qc_metrics, the value will be an object containing various QC metrics. @shihabdider will provide updatedmetadata.jsons for the clinical samples by end of Tues Oct 15.Threshold Specification:
@kevinmhadi will provide the specification of the QC metrics object's attributes and their respective thresholds (in this issue thread) to indicate bad samples. For now these will be flat comparisons, later we can decide on more sophisticated criteria for flagging.
Distribution Plot:
@kevinmhadi will also specify which attributes should be aggregated across patients to generate distribution plots for them.
JaBbA QC: Not sure how we this information to be presented. One possibility is just as an additional (optional) attribute of this qc tooltip (in which case we can rename the parent attribute to
sample_qc_metricsand include it as part of the metadata.json). Alternatively we can try to incorporate it into the JaBbA plot component somehow (maybe an indicator next to the title that doubles as a tooltip and only appears when there's something wrong?). @jrafailov if you have some other ideas please share in this thread. In particular, we need to decide what file to put this information (or introduce a new file) that will be fetched by the case-reports front-end.