Pick a precursor m/z, the MS2 scans for it are averaged, and SIRIUS is asked which molecular formula could produce that fragmentation.

How to use it

  1. Upload one or more .RAW files. Precursors are pooled across all of them.
  2. Choose a precursor m/z. Exhalation windows are auto-detected from the TIC; check them on the TIC plot and use Modify windows if they are wrong.
  3. Click Fit & validate under Mass calibration. It reports the --ppm-max this file supports — put that in the MS1 ppm box. If it says NO_GAIN, the file is already calibrated and nothing is changed.
  4. Ionization is read from the file's scan filters, not chosen — check the adducts offered for that polarity, then Run SIRIUS Analysis.
  5. In Results, click a feature → a formula → a structure. Each click drives the panels below it.

What the scores mean

SIRIUS score Overall support for a molecular formula. Compare candidates within one precursor; the gap to #2 says more than the value.
Tree score How well the fragmentation tree explains the peaks — each peak given a fragment formula consistent with the parent, linked by plausible neutral losses.
Explained % Share of the measured spectrum the tree accounts for. Read it together with the tree score, not alone.
CSI:FingerID A second step, run per formula: it predicts a molecular fingerprint (thousands of yes/no substructure flags) from the spectrum, then ranks database structures by how well they match. Ranks structures within one formula.
COSMIC Confidence that the top CSI:FingerID structure is actually correct. The closest thing here to a probability.
CANOPUS Predicted compound class (e.g. 'fatty acyls') straight from the spectrum — no database entry needed, so it often works when the exact structure does not.
Δ ppm Candidate's theoretical mass against the precursor. Weak on its own — every candidate already passed the search's mass filter — and only meaningful once the mass is corrected. Which mass it was computed from is stated under the table, on the line beginning “measured…” or “isolation setpoint…”: lock-mass corrected, UNCORRECTED (drift on these files reaches ±13 ppm — see the Mass calibration tab), or the isolation setpoint rather than a measurement.
P(iso) How well the candidate explains the measured MS1 isotope envelope — the one column carrying evidence SIRIUS never saw, since the search uses MS/MS and the precursor mass only. Every Full-MS1 scan in the file is averaged, and each candidate's own predicted isotope peaks (¹³C, ¹⁵N, ³⁴S, ³⁷Cl, ⁸¹Br …) become windows to look in.
Higher is better, and it is normalised over the candidates listed, so it sums to 1 even if the right formula is not among them. Read it as a ranking within this list, never as a probability of being correct.
Near-equal values mean the envelope cannot separate those candidates — the usual case below ~m/z 150, and the honest answer rather than a failure. A value near 0 is the informative one: that candidate predicts a peak the file is sensitive enough to have seen, and it is not there.
Blank only when the file cannot support the measurement at all — no full-scan MS1, or no MS1 peak within 5 ppm of the precursor.
HMDB How many known structures carry the formula. A hit is weak evidence and a miss almost none: many breath volatiles are exogenous or microbial and absent from a human-metabolome database.
Vol. Volatility class from the saturation mass concentration C* (µg m⁻³, 300 K), predicted from the atom counts by the Li et al. parameterisation: log10 C* = (nC0 − nC)bC − nObO − 2(nCnO/(nC+nO))bCO − nNbN, with per-class parameters for CH, CHO, CHN and CHON. Bins are the standard VBS ones: VOC log10 C* > 6.5, IVOC 2.5–6.5, SVOC −0.5–2.5, LVOC −3.5 to −0.5, ELVOC < −3.5. Only VOC and IVOC are plausibly in the gas phase at the inlet. Blank means the formula carries an element outside CHON (S, P, halogen), which the parameterisation has no terms for — a blank is 'not estimated', not 'not volatile'. The number itself is printed under the candidate table. This is a formula-level estimate, so isomers cannot differ.

The tabs

  • MS2 Scans — scans for the selected precursor, plus its XIC with the exhalation windows shaded.
  • Mass calibration — measured mass error before/after correction, which drift model was chosen, and the ppm tolerance the file supports.
  • Fragment modulation — each fragment's behaviour during exhalation vs non-exhalation, to show which peaks in the isolation window belong together.
  • Results — formulas, structures, fragmentation tree, mirror plot, neutral losses, and the HMDB structures sharing the chosen formula.

Acquisition overview

Everything below describes the full-MS (MS1) survey scans of the active file only.

Extracted Ion Chromatogram

Mass error vs m/z — before and after correction

Points are individual scans, the line is each lock mass's median ± MAD. An anchor marked × was judged inconsistent with the others and left out of the fit — it is drawn so you can see why.

What the instrument itself did

Whether the deviations above are the full mass error or only what the instrument's own lock mass left behind. Read from the file's per-scan Trailer Extra record, not from the method notes.

Model comparison (leave-one-out |p99|, ppm)
Per-anchor screen

            

            

              
Per-fragment behaviour across the breath cycle

              
              

              

              

              
What the MS/MS looks like with and without the background fragments

              Left: the averaged MS/MS as that row's procedure produces it. Right: the same spectrum after the breath-phase filter — the identical call the run makes, so it removes exactly the peaks the run would remove and nothing else. Only red (background-depleted) peaks are removable; grey (flat) peaks are neither breath-enriched nor background and always survive. Top row is custom averaging, bottom row Thermo's — both are built every time so they can be compared. Hover anywhere to read off the nearest peak's m/z and intensity. Preview only; no plot here changes the search.
              

                

                  

                

                

                  

                

              

              

                

                  

                

                

                  

                

              

              

              

            

            

              

                

                  

                    

                    
Candidate formulas  (click a row)

                    

                    

                    
Known structures with this formula (HMDB)

                    

                    

                    

                    
Candidate structures (CSI:FingerID)  (click a row)

                    

                    

                    

                    

                      

                      
De novo structures (MsNovelist)

                      

                    

                    

                    
Fragmentation tree

                    Interactive: scroll to zoom, drag to pan, hover a node for the full annotation.
                    

                    

                    
                      
                      Download as PNG
                    
                    

                  

                  

                    
Annotated MS/MS mirror plot

                    

                    
                      
                      Download fragmentation data (CSV)
                    
                    

                    
Explained vs unexplained peaks

                    SIRIUS-style spectrum: peaks explained by the fragmentation tree are teal, unexplained peaks grey.
                    

                    

                    
Neutral loss plot

                    

                      
                      

                        
                        
                      

                    

                    

                    
                      
                      Download neutral-loss data (CSV)
                    
                    

                    
CANOPUS compound classes

                    

                  

                

              

              

                

                  
No results yet

                  
Upload a RAW file and run the SIRIUS analysis.