ASMS Fall 2018 Metabolomics Informatics Workshop Peak Picking

1. 1 Principles of Peak Picking and Alignment Emma L. Schymanski FNR ATTRACT Fellow and PI in Environmental Cheminformatics Luxembourg Centre for Systems Biomedicine (LCSB), University of Luxembourg Email: emma.schymanski@uni.lu …and many colleagues who contributed to my science over the years! ASMS Fall Meeting, San Francisco, California, November 29-30, 2018 Image©www.seanoakley.com/ https://tinyurl.com/asmsfall2018-peaks How many peaks will a peak picker pick if a peak picker only picks peaks?

2. 2 (nevertheless, I will do my best!) DISCLAIMER! MS1 MS2 Two very different worlds …

3. 3 Presenting Peak Picking: Plan o Why Peak Pick o Terminology • Peak Picking vs Centroid vs Profile … o Peak Picking & Peak Pickers • “best of” xcms and enviPick • Peak Picking in Pictures • Peak Picking Parameters • Alleviating Peak Picking Parameter Panic o Alignment ( / Profiling) • “best of” xcms and enviMass o Peak Picking Pointers o Don’t just listen to me … do it!

4. 4 Why Peak Pick (I) Example scheme of liquid chromatography - mass spectrometry Image © www.planetorbitrap.com/q-exactive Sampling Extraction (SPE) HPLC separation HR-MS/MS

5. 5 Why Peak Pick (II) This is what the output “really” looks like … Image © www.planetorbitrap.com/q-exactive

6. 6 Why Peak Pick (III) Identification = turning numbers into structures N N N S CH3 NHNH CH3 CH3 CH3 N N N S CH3 NHNHCH3 CH3 OH P O S SO CH3 CH3 CH3 P OHS S O CH3 CH3 OH CH3 S O O OH CH3 CH3 S N S O O OH S O O OH CH3 CH3 S O O OH CH3 CH3 S O O OH CH3 CH3 S O O OH CH3 CH3 S O O OH CH3 CH3 N N N S NHNH CH3 CH3 CH3 NH2 OH O massbank.eu

7. 7 TERMINOLOGY! o Peak picking can be multi-directional, i.e. • in mass… or time…

8. 8 Mass: Centroid vs Profile Data (enviPat) https://www.envipat.eawag.ch/index.php and Loos et al Anal. Chem. 87(11), 5738-5744. DOI: 10.1021/acs.analchem.5b00941

9. 9 Mass: Centroid vs Profile Data (enviPat) https://www.envipat.eawag.ch/index.php and Loos et al Anal. Chem. 87(11), 5738-5744. DOI: 10.1021/acs.analchem.5b00941

10. 10 TERMINOLOGY! http://proteowizard.sourceforge.net/ o Peak picking can be multi-directional (mass, time) • Peak picking in Proteowizard MSConvert is “centroiding” masses (turning profile mode data into centroided data for efficient processing)

11. 11 Peak Picking (in time) Source: R. Tautenhahn, C. Böttcher, S. Neumann, BMC Bioinformatics 2008, 9:504. DOI: 10.1186/1471-2105-9-504 o Peak picking along time axis (chromatographic peaks)

12. 12 Peak Picking Source: R. Tautenhahn, C. Böttcher, S. Neumann, BMC Bioinformatics 2008, 9:504. DOI: 10.1186/1471-2105-9-504 o Peak picking along time axis (chromatographic peaks)

13. 13 Peak Picking Source: Johannes Rainer; http://bioconductor.org/packages/release/bioc/vignettes/xcms/inst/doc/xcms.html o Peak picking along time axis (chromatographic peaks)

14. 14 Peak Picking Source: Johannes Rainer; http://bioconductor.org/packages/release/bioc/vignettes/xcms/inst/doc/xcms.html o Peak picking along time axis (chromatographic peaks) Several Samples Overlaid Red = KO Blue = wild type Rectangle = chromatographic peaks identified per sample

15. 15 Peak Picking o Several options for peak picking • XCMS and centWave • Tautenhahn et al 2008 DOI: 10.1186/1471-2105-9-504 • http://bioconductor.org/packages/xcms/ • MZmine 2 • Pluskal et al 2010 DOI: 10.1186/1471-2105-11-395 • http://mzmine.github.io/ • enviPick / enviMass • Loos 2018 DOI: 10.5281/zenodo.1213098 • http://www.looscomputing.ch/eng/enviMass/overview.htm • Plenty of other open, research and vendor options ...

16. 16 Peak Picking o Result is something like this (from Formulator output):

17. 17 Peak Picking – XCMS & XCMS Online o http://bioconductor.org/packages/xcms/

18. 18 Peak Picking – XCMS & XCMS Online o https://xcmsonline.scripps.edu/

19. 19 Peak Picking – enviMass and enviPick o http://www.looscomputing.ch/eng/enviMass/overview.htm o R packages …

20. 20 Peak Picking in Pictures http://www.looscomputing.ch/eng/enviMass/topics/peakpicking.htm Red = peaks Grey = noise

21. 21 Peak Picking .. Somewhat simpler picture http://www.looscomputing.ch/eng/enviMass/topics/peakpicking.htm

22. 22 centWave – Gaussian with “Mexican Hat” https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-9-504

25. 25 But … peaks are not perfect! http://www.looscomputing.ch/eng/enviMass/topics/peakpicking.htm o See enviMass website for explanation …

26. 26 Critical Point: Separating Peaks from Baseline

27. 27 Peak Picking Parameters https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-9-504 o There are a lot of options to tweak! • I will just run through (main) centWave parameters • enviPick is too complicated => further reading!

28. 28 Peak Picking Parameters: centWave ppm maximal tolerated m/z deviation in consecutive scans, in ppm (parts per million) NOTE: dependent on your mass spectrometer

29. 29 Peak Picking Parameters: centWave peakwidth Chromatographic peak width, given as range (min,max) in seconds NOTE: highly dependent on your chromatography!

30. 30 Peak Picking Parameters: centWave snthresh Signal to noise ratio cutoff

31. 31 Peak Picking Parameters: centWave prefilter prefilter=c(k,I). Prefilter step for the first phase. Mass traces are only retained if they contain at least k peaks with intensity >= I Only one “stick” so will fail recommended prefilter settings

32. 32 Too Many Peak Picking Parameters ??????? https://bioconductor.org/packages/ release/bioc/vignettes/IPO/inst/doc /IPO.html o IPO to the rescue! o Parameter optimization for xcms-based workflows … o Libiseller et al 2015, DOI: 10.1186/s12859-015-0562-8 IPO = Isotopologue Parameter Optimization

33. 33 Too Many Peak Picking Parameters ???????

34. 34 RECAP: Why Peak Pick? Identification = turning numbers into structures N N N S CH3 NHNH CH3 CH3 CH3 N N N S CH3 NHNHCH3 CH3 OH P O S SO CH3 CH3 CH3 P OHS S O CH3 CH3 OH CH3 S O O OH CH3 CH3 S N S O O OH S O O OH CH3 CH3 S O O OH CH3 CH3 S O O OH CH3 CH3 S O O OH CH3 CH3 S O O OH CH3 CH3 N N N S NHNH CH3 CH3 CH3 NH2 OH O massbank.eu

35. 35 o Instruments change over time … o Before we can do fancy statistics, we need to make sure our samples are comparable!

36. 36 Alignment http://bioconductor.org/packages/release/bioc/vignettes/xcms/inst/doc/xcms.html#3_initial_data_inspection o Alignment / Profiling => which peaks belong together across large sample sets?

37. 37 Alignment http://www.looscomputing.ch/eng/enviMass/topics/profiling.htm o “Profiling” in enviMass

38. 38 Alignment ~= Retention Time Correction http://bioconductor.org/packages/release/bioc/vignettes/xcms/inst/doc/xcms.html#3_initial_data_inspection o Many algorithms and methods … o Before:

39. 39 Alignment ~= Retention Time Correction http://bioconductor.org/packages/release/bioc/vignettes/xcms/inst/doc/xcms.html#5_alignment o Many algorithms and methods … o After (Obiwarp algorithm in xcms)

40. 40 Before Alignment After Alignment

41. 41 Changes over samples http://bioconductor.org/packages/release/bioc/vignettes/xcms/inst/doc/xcms.html#5_alignment o Difference between adjusted and raw retention times along the retention time axis

42. 42 Some advice … o Peak pickers are designed to pick the perfect peak • But life is never perfect and peaks are no different o Pick the peak picker that is best for your situation • Convenience, ease of use, designed for your data, … • The optimal choice is usually a compromise o Be sceptical (visualise your data, reality check it, etc.) • But don’t go overboard in evaluating peak pickers … remember your (real) goal …

43. 43 Peak Picking Overlap (centWave paper) https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-9-504

44. 44 Verify with EIC Extraction [these are NOT picked] https://github.com/schymane/ReSOLUTION/blob/master/R/RMB_EIC_prescreen.R No peak at all Nice peak, MSMS Peak, no MSMS Noise with MSMS (careful!) Isobars with MSMS (careful!)* Looking for chemicals known to be present in the sample

45. 45 Just because you find a peak … ENTACT Project: https://www.epa.gov/sites/production/files/2018-06/documents/comptox_cop_6-28-18.pdf o Mix 505: One candidate with this mass/formula • DTXSID9040001, C9H8O4 o One chemical… How many peaks?

46. 46 …doesn’t mean it’s your compound of interest!

47. 47 Beware of artefacts! o Your results also depend on the acquisition data!

48. 48 Further reading DOING! [Vendor independent] o Don’t just take my word for it … don’t just read about it … DO IT. There are so many ways to try it out … complete with sample data! [Open Science!] o http://bioconductor.org/packages/release/bioc/vignettes/x cms/inst/doc/xcms.html o http://www.looscomputing.ch/eng/enviMass/overview.htm o An interface that many enjoy, likely comes with example data but requires a login … o https://xcmsonline.scripps.edu/

49. 49 Further reading DOING! [Vendor independent] o http://mzmine.github.io/ o http://prime.psc.riken.jp/Metabolomics_Software/MS-DIAL/ o MS-DIAL

50. 50 Acknowledgements emma.schymanski@uni.lu Further Information: http://bioconductor.org/packages/xcms/ http://www.looscomputing.ch/eng/enviMass/overview.htm https://xcmsonline.scripps.edu/ http://mzmine.github.io/ EU Grant 603437 The CompMS Community (proxy photo)

51. 51 Extra Slides

52. 52 Quality Control of Data Slide c/o Michael Stravs o Always visualise results … never take anything for granted

53. 53 Homologues: Challenge Peak Pickers but are Present! Stravs et al. (2013), J. Mass Spectrom, 48(1):89-99. DOI: 10.1002/jms.3131 OHSO O CH3 O OH m n SPA-9C m+n=6 www.massbank.eu ACCESSIONS (LAS, SPACs): Literature MS/MS LIT00034, LIT00037 Std Mix., Sample ETS00012, ETS00018https://github.com/MassBank/RMassBank/ Tentatively Identified Spectra: http://goo.gl/0t7jGp

54. 54 Be wary of instrument specific phenomena! o R package nontarget: satellite peak removal

55. 55 Be wary of instrument specific phenomena II o Orbitrap-specific calibration issues (not observed in TOF)

ASMS Fall 2018 Metabolomics Informatics Workshop Peak Picking

Emma Schymanski

ASMS Fall 2018 Metabolomics Informatics Workshop Peak Picking