Yesterday I made a short post on this new study in Open Reports. I'll be honest, I thought it was a seriously impressive sample set, but I was a little skeptical about the results. Not that I didn't believe the authors could find the bacterial proteins, but I was skeptical that the results could be that clean cut. I've seen a few "identification of digested bacterial proteins in body fluid" studies and even with high resolution data the matrix is complex enough that there is noise there. These results just seemed a little too good to be true.
Shoutout to ProteomeXchange/PRIDE and these authors for making this data extremely easy to obtain, sort and reprocess overnight!!
For my re-analysis, I threw all the files into PD 2.2 and used the default processing workflow for LTQ-Orbitrap LFQ and LFQ concensus with enhanced annotation. I made a 77MB FASTA in PD by parsing a UniProt TrembL I downloaded in July on "Steptococcus pneumoniae" and used my normal Uniprot SwissProt human + cRAP databases.
I can confidently say, with no reservations at all, that I believe this group did exactly what they said they did. This readout is the simplest one to display what I'm looking at. Green means protein found in that sample.
I have to go into really low scoring stuff before I see a peptide or two from the bacteria flagged as present in the controls -- and the first 2 match cRAP hits as well.
Okay...so this study wasn't published for almost a year from the time it was submitted. I'm willing to bet $10 that part of the holdup was that a reviewer (or two) didn't believe the data could be THIS clean. It is.
I don't know how long it takes to culture CSF in a classical microbiology assay to test for these bacteria -- or how trustworthy results from those classic assays are. What I do know --this team has shown clearcut evidence that our technology can make determinations of CSF infection with 2 hours of instrument time per patient (on a mass spec first released 7 years ago!)