1. DIA: the Why, How, and When…Really…

2. Outline: 20 Minutes to Clarity
Why?
How?
When?
In case the title slide wasn’t clear enough. This is a very fundamental intro for people who have never done DIA, compiled by someone who’s never done DIA, from information shared from people who have and know what they’re talking about…

3. WHY: What’s all the fuss about?
Potential to quantify all detectable peptides with high sensitivity
Continuous collection of MS2 spectra across peak for whole mass range
Qualitative AND quantitative results
All product ions, all precursors – integrate and/or search against library
Minimal up front method development (relative to SRM)
Same data acquisition method applicable to all sample sets, all sample complexities
Selectivity can be optimized post-acquisition to offset sample complexity
More transitions with proper relative ratios  more selectivity

4. Sounds AMAZING… What’s the catch?
In a word… Interference.
Intensive post-acquisition data processing (software-dependent), manual target validation, quantitative error modeling, and statistical analysis required
Spectral libraries (preferably with retention time markers) required for qualitative analysis and greatly facilitate quantitative analysis
ie, results from DIA experiments will be dependent on the quality of your prior deep-dig DDA analysis of sample type and spectral library creation
Reproducible chromatography and retention time characterization/ calibration critical for selectivity and validation
Not compatible with fast chromatography
as duty cycle increases, so must average peak width
efforts to improve selectivity at MS method level result in longer cycle times, in turn requiring further “dumb-ing down” of chromatography
Mass accuracy & precursor range isolation width have huge impact on results
by nature, this method introduces reproducible interferences
Very large raw files generated

5. HOW: Recommended DIA Workflow and Data Processing
Biological Fluid Samples
The workflow at a glance… its really important to understand that without spectral libraries, you will be doing a LOT of manual validation. (Even with spectral libraries, you should do manual validation, but it’ll be automated through Pinpoint/Skyline). It should also be noted, you’re still limited to what you can ID in a DDA experiment, unless you go completely hypothesis-driven and look for all fragments of all charge states of all peptides of interest from all proteins of interest….
DIA experiment
DDA Experiment
Targeted protein selection with Proteome Discoverer
Spectral library generated with Pinpoint/Skyline based on PD Search results
Validation, Normalization, &
Statistical Analysis
Targeted data extraction
(Retention time, m/z & fragment ion distribution)

6. Do I really need a spectral library?

7. Spectral Libraries Improve DIA Data Interpretation
Spectral Libraries provide information on the following:
Which peptides per protein have been previously identified from a true tandem mass spectrum
What is the relative abundance of the specific peptides to others from the targeted protein
Determining the precursor charge states for each targeted peptide
Determining the product ion distribution – which fragments should be seen
Determining the relative retention time for each targeted peptides
If no spectral libraries, how to increase effectiveness?
Increase the mass tolerance (i.e. tighten ppm values used for data extraction)
Incorporate an alternative approach for correlating peptide sequence with measured retention time – SSRCalc, PRTC peptides

8. Generating a Spectral Library in Pinpoint
Spectral library is generated from imported PD results
How to import PD results (.msf file) into Pinpoint to utilize as spectral library
Only peptides identified with good quality ms/ms spectra are used for the spectral library
**Coming soon… Crystal node in PD for more statistically rigorous generation of custom spectral libraries**

9. DIA Assay Development Based on Spectral Library
Up to eight fragment ions used for sequence confirmation
Most intense three fragment ions used for quantification

10. Comparing the Library Spectrum with DIA Data
DIA data can be uploaded into Pinpoint or Skyline for comparison against spectral library…
Dot product correlation coefficient of 0.91
P-value score of 1.45e-3

11. Processing DIA Data Without Spectral Libraries
Targeting KGNVATEISTER without knowledge of product ion distributions using ± 25 ppm
This slide demonstrates the difficulty in determining which peak to assign to a peptide WITHOUT spectral library information. Both peaks look reasonable…

12. Comparative Integration Times for FVTQAEGAK
5.85 minutes
8.42 minutes
When fragment ion intensities are compared to a library spectrum, the choice becomes much more clear. This also demonstrates the importance of retention time characterization of data. Use of RT calibration standards add extra evidence to choice of peak to integrate.
Spectral libraries provides an indication of product ion distribution – which fragments should be more abundant as well as provides additional means of verification through product ion distribution overlap.

13. Retention Time Landmarks in Library Improve Robustness of Target Identification in DIA Experiment
Sample PRTC Peptides 30 40 50 60 70 80 90
100
110
Spiking the PRTC kit into a sample provides RT landmarks from which the data analysis can establish reference points for all endogenous targets can be mapped against. The addition of the PRTC kit facilitates a constant set of “knowns” that have expected LC behavior profiles regardless of gradient, gradient length, or specific C18 packing material.
The PRTC kit facilitates direct RT correlation between different experiments – as shown in the figure. Each peptide is evaluated by measured RT values for the two experiments, as is done for all PRTC peptides. The resulting RT ratios across experiments can then be plotted as a function of one of the experimental gradients.
Once the correlation equation is calculated, then any endogenous peptide identified in the library, the empirically determined RT value can then be substituted into the PRTC RT equation to calculate the expected RT value in the new experiment. 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44
Library PRTC Peptides

14. Adding in Retention Times for Further Confirmation
50 ppm
The retention times from the libraries were determined for each targeted peptide based on a true tandem MS and matched values. Using these RT values locally (specific peptide) means nothing, but comparing against all other targeted peptides builds confidence.
DIA Experimentally Determined Retention Time (min)
10 ppm
Library Retention Time (min)

15. OK fine, I’ll make good libraries
OK fine, I’ll make good libraries. Now about collecting the data, and turning it into reliable information…

16. Internal Standards
Data needs to be normalized across all files (replicates and experiments) to ensure detection of TRUE biological changes. This is not trivial!
Inclusion of internal standards (N15 labeled sample, synthetic heavy labeled target peptides) makes post-acquisition normalization and data interpretation much more straightforward

17. Collecting Your DIA Data…Pick Your Poison:
Basic DIA (repeated large isolation windows)
Lower selectivity, more interference, easier to set up
0files/tips/SkylineDIAMiniTutorial_2_1.pdf (DIA & msxDIA)
MsxDIA
Higher selectivity due to randomized, smaller, multiplexed isolation windows
More sophisticated processing (requires de-multiplexing) available in Skyline only
See Skyline link above
WiSIM DIA
Available on Fusion only, version 1.1 of software
Novel workflow based on high res MS1 quant, low res MS2 verification
Higher throughput, less interference, easier to set up and process
Now you know how to generate your library for analysis, how do you want to collect your DIA data? Here are a few flavors of DIA… More are emerging, including overlapping window DIA and variable isolation window DIA, all to improve selectivity/ reduce interference issues…however these will require software support that’s not quite there. This is what you can do now (or, almost now, pending Fusion 1.1 release)
MsxDIA are small 4 Da windows. You must have Skyline.

18. Impact of Precursor Isolation Window Width
Precursor isolation window width has the largest impact on reducing interference.

19. Effects of Resolution Setting for DIA Acquisition
You can see here that resolution setting does have some effect on the number of accurately assigned peptide targets as a result of resolved interference, however…

20. The Interplay of Isolation Window and Extraction Window
SWATH/XIC
DIA trap
MSe/AIF
Combined importance of MS method isolation width and post-acquisition processing m/z extraction tolerance. Intuitively, the tightest combination of isolation and extraction will provide the best results WITH RESPECT TO REDUCING INTERFERRENCE. This set of conditions however is un-realistic, as the duty cycle for a DIA method with 1Th isolation windows across the m/z range would be monstrous, and not applicable to chromatography. Interestingly, the ability to extract with high m/z tolerance (10ppm) enables the use of a much larger isolation window (to maintain duty cycle), while still retaining a significant # of interference-free transitions for most peptides, demonstrating the importance of high mass accuracy.

21. Simultaneous Qualification and Quantification Using Pinpoint
Final DIA data output in Pinpoint. Pinpoint weights the averages based on relative abundance – more intense transitions, isotopes, peptides, etc. contribute more to the overall protein response. It has also added in the capabilities of performing both global and local normalization. Like any other quantitative software program, it is in the researchers best interest to validate quantitative output (averages and error propagation).
Qual using eight most intense fragment ions
Quan
out put
Quan using
three most intense
fragment ions

22. Comparison of DIA Processing Software
Skyline is another great (and free) option for processing DIA data with great online tutorials and support.

23. Comparison of DIA Processing Software
Pinpoint
Skyline
Spectronaut
Requirements
No special requirements
Standard Calibration Kit (see Con below)
Data format conversion
Spectral library
Pro
.msf file can be imported directly as the spectral library
Thermo specific software supporting all targeted proteomics experiments, including SRM, PRM, targeted DDA, and DIA experiments
Can filter out peptides by using FDR filter in the .msf file while importing the .msf file
Using RT from the spectral library for peptide filter
Direct view of the product ion distribution vs. spectral library
CV% and Calibration curves are automatically calculated
Free Open-Source Windows client for building SRM, MRM, PRM, targeted MS/MS, targeted DDA and DIA/SWATH, as well as msxDIA
Very stable and be able to handling big data sets in one shot
Be able to generate inclusion mass lists for DIA experiments
.msf file from P.D.1.4 can be imported directly as the spectral library
The peptides in the .msf file must be pre-filtered before importing
Fast Developments and Features implementation
Free academic license
Supports Q Exactive (and TripleTOF 5600 & 5600+)
Application Note on QE - quantitative proteomics discovery. Comparison of DIA and Shotgun techniques
Using FDR for peptide validation
Con
Does not de-multiplex msxDIA data
Have difficulty to handle big data
sets (facing crashes while loading QE data sets in one shot)
Many settings in the workbook
are not saved after iterations (still some SW bugs)
RT from the library is not used as peptide
validation factor yet
2. The peptide validation score is still not very
understood
1. .raw data files have to be converted into
.htrms file
- Conversion time is substantial
2. For small molecules Spectronaut is not
the solution
Standard Calibration Kit :
a peptide mix comes at a cost to customer

24. And this really works even better on the Fusion???

25. WiSIM-DIA Experiment on the Orbitrap FUSION
How it works
Three high-resolution, accurate-mass (HR/AM) selected ion monitoring (SIM) scans (240,000 FWHM) with wide isolation windows (180 amu) were used to cover all precursor ions of 450 – 990 m/z. In parallel with each SIM scan, 15 sequential ion trap MS/MS with 12 amu isolation windows were acquired to cover the associated 180 amu SIM mass range.
Orbitrap
R:240,000
Linear
Trap
SIM
amu
amu
amu
----
15 sequential cid ms/ms
scans with 12 Da
isolation Windows
456
m/z
468
600
612
624
636
648
780
792
804
816
828
960
972
984
Instrument method template will be available in OT Fusion 1.1

26. WiSIM-DIA Data Processing
Pinpoint 1.3 software
A spectral library is established using PD search results from the short gun data dependent discovery experiments.
For quantification, the XICs of isotope C12 and C13 precursor ions per targeted peptide are extracted from the HR/AM SIM data with ± a 5 ppm window.
For peptide sequence confirmation, eight most intense fragment ions (b and y types) detected from discovery data are extracted from cid ms/ms with ± 600 ppm window and used to match the spectral library.
A peptide with a P-value of correlation with library spectra that is less than 0.1 is considered to be confirmed with high confidence by the spectral library match.

27. Establishing a Spectral Library
Spectral library is generated from imported PD results
Important here, as before, you want to build a spectral library from the same instrument… MS2 spectra will be low res for this method, so use low res MS2 data to build the library.
Only peptides identified with good quality of ms/ms spectra are used for the spectral library

28. Targeted Assay Development Relying on Spectral Library
Up to eight fragment ions used for sequence confirmation
Isotope C12 and C13 precursor ions used
for Quan

29. Simultaneous Qualification and Quantification Using Pinpoint
out put
Qual using eight most intense fragment ions
Quan using
Isotope C12 and C13 precursor ions

30. Success of Targeted Peptide Quantitation in HeLa Cell Lysate Digest
Yes, MS1 HR/AM quant is the best.

31. Advantages of WiSIM-DIA
By using precursor ions collected on SIM with extremely high resolving power of 240,000 for quantification, high sensitivity and high selectivity are achieved through separation of most background interferences from analyte signal.
• Down to 10 attomole LOD/LOQ
• Up to 4 orders of linear dynamic range
Capability to quantify peptides which have poor fragmentation efficiency, such as large peptides, labile peptides and peptides with PTM.

32. WHEN: Areas of Application for DIA
Sample from immuno-precipitation for protein-protein interaction studies (network topology) (Anne-Claude Gingras Mount Sinai, Toronto)
PanOmics studies to elucidate enzymatic pathways (medium to high abundance proteins) (Ruedi Aebersold, ETH, Zürich)
Simplified and enriched PTM fractions, including glycans
Comprehensive coverage of biopharmaceutical peptide maps from enriched growth media (MSe Waters apps note)
Small molecule applications in metabolomics, clinical, food and environmental markets in need of comprehensive sample coverage
Finally, where does DIA have a legitimate place if experiments are planned properly?
Here are a list of applications that have been demonstrated in the literature already. What do these applications have in common?
Reduced complexity: hundreds of proteins rather than thousands of proteins (Plasma is not a good sample)
Reduce dynamic range, typically between 2-3 orders of magnitude (for biomarker discovery in plasma SWATH is not a good method)
Applications that benefit from a comprehensive capture of analytes in the sample. (affinity purification)

33. Background slides… Info courtesy of… Scott Peterman Andreas Hühmer
Reiko Kiyonami
Benjamin Orsburn
Lani Cardasis

34. DIA Data Processing Pinpoint 1.3
Spectral library established using DDA discovery data
Simultaneous peptide sequence confirmation and quantification
Three most intense fragment ions (b and y types) for quantification
Eight to ten most intense fragment ions (b and y types) for confirmation
Automatic quality control of transitions to eliminate transitions with significant interferences
± 5 ppm XIC window

35. Adding in Retention Times for Further Confirmation
50 ppm
The retention times from the libraries were determined for each targeted peptide based on a true tandem MS and matched values. Using these RT values locally (specific peptide) means nothing, but comparing against all other targeted peptides builds confidence.
DIA Experimentally Determined Retention Time (min)
10 ppm
Library Retention Time (min)

36. Building in Targeted Peptide Lists
Build acquisition method
Acquire data
Assign m/z values to each peptides
Process data
Qualitative analysis
Quantitative analysis
Select peptides/proteins
DIA
1. Biological hypothesis used to select targeted proteins
2. Determine peptides per targeted protein
Do you have spectral libraries?
Spectral Library No
Yes
Build target list based on in silico assumptions
Generate all possible peptides
Assume most likely precursor charge states
Predict product ions based on sequence
Predict retention times based on hydrophobicity factors
Build target lists based on spectral library information
Peptide sequences
Precursor charge states
Most abundant product ion m/z values
Product ion distribution values
Retention time values
Lastly – peptide selection is driven by the biological hypothesis – if determining the protein level, any unique peptide will do, if the assay is driven by site-specific targeting (i.e. SNPs, PTMs, truncation) then the peptide covering this site must be targeted.
Courtesy of Scott Peterman

37. Target Peptide Verification
Build acquisition method
Acquire data
Select peptides/proteins
Assign m/z values to each peptides
Process data
Qualitative analysis
Quantitative analysis
DIA
Accurate mass values for each precrusor/product ions
Extract product ion XICs using a specified mass tolerance (theoretical m/z value +/- set mass window e.g. 10 ppm)
Overlaid XICs for multiple mass values per targeted show co-variance based on user- defined mass tolerance values
Measured retention times align with either spectral library retention times or other means of predicting retention times (SSRCalc)
Determine the AUC values per product ion and calculate product ion distribution for comparison to library/expected distribution
Courtesy of Scott Peterman

38. Importance of Mass Accuracy for Data Interpretation
Y = x –
R2 =
What is the extraction mass tolerance effects for large data sets? From this plot, using 10 ppm extraction tolerance and the product ion distribution, Pinpoint was extremely successful in identifying the correct RT over a 70 minute processing window. There are a few that fall off of the line but the linear regression was
YYWGGQYTWDMAK
Courtesy of Scott Peterman

39. Importance of Mass Accuracy for Data Interpretation
Courtesy of Scott Peterman

40. Importance of Mass Accuracy for Data Interpretation
YYWGGQYTWDMAK
Same approach as shown in slide 37 except the mass extraction tolerance was increased to 75 ppm to simulate AB’s processing approach (see slide 4). The product ion rank order as defined by the spectral library was still used as was in slide 17 but the results show greater difficulty in identifying the correct RT peaks. There is one peptide highlighted and shown in Slide 20.
Courtesy of Scott Peterman