qcqa via scan level metrics

[rmflight]

Would be great to have examples of this. The ones I can think of that we've seen are:

• [x] Failing b/c we don't get enough scans for normalization (seeing this now in rerun of NSCLC)
• [x] Scans being dropped b/c their sqrt terms are outliers
• [x] Scans being dropped due to bad R^2 (can check if this happens I think)
• [x] HighSD is an example (already have this in the context of HPD sites)
• [x] Two sets of different start - stops on range of convertable frequency differences
  • We saw this in some NSCLC samples, and can actually show it, and how we modified the scan filter to remove the scans.

[rmflight]

Another way to check for scans being dropped for any reason would be to open the zip, do the raw_ms filtering, and then compare the normalization factor scans from the final zip to the scans in the raw_ms filtering. Or even what is listed in raw_ms$scan_range and what is in the normalization factor data.frame.

[rmflight]

Scans being dropped due to bad R^2 (can check if this happens I think)

Based on the logs for re-running the entire set again, we don't see this come up. So we can mention it, but we don't see it happen.

Scans being dropped b/c their sqrt terms are outliers.

We missed this one above, but it does happen. Many, many samples lose one or two scans due to having outlier sqrt terms coefficients in the frequency model. What is really nice, is we can see how many samples and how many scans were lost due to this, and plot a sina_plot with the outlier terms. I've got the data for this.

Two sets of different start - stops on range of convertable frequency differences.

I need to pull this data in and make the graph, and note how many samples this happened to and how we got around it.