m. mcbrien, a. galin, a. vazhentsev, and vitaly lashin ... · ymc pro c18 ph 2.5 xterra rp 18 c18...
TRANSCRIPT
Advanced Chemistry Development, Inc. (ACD/Labs)
n-Dimensional MAP Chromatography:
Virtual Resolution of Components Based on
Orthogonal Chromatographic Methods
M. McBrien, A. Galin, A. Vazhentsev, and Vitaly Lashin
Pittcon – February 25/07
2
Outline
Why match peaks?Reviewing MS-MAPOrthogonal separations and chromatographic screeningPulling it all together – NDMC and “virtual resolution”Limitations of the MAP algorithmsThe future
3
Typical Chromatographic MethDev
Optimization of continuous variables –i.e., concentration of organic modifierSelectivity changes are modest; typically you are “tweaking” a promising method
4
Automated Method Development? The Challengetfa series.esp
605550454035302520151050Retention Time (min)
12
34
56
78
910
1112
1314
1516
Go from a series of collected datafiles under different conditions…
5
Automated Method Development? Peak Matching
90858075706560555045403530
Solvent B, %
6
8
10
12
14
16
18
TFA
, mM
3.893.643.403.162.912.672.432.191.941.701.461.210.970.730.490.240.00
…to a model of the sample’s behavior over any value of those conditions:
6
Peak Matching
Perhaps the greatest challenge in computer-assisted method development.All peaks must be tracked across all experimentsHumans use spectra, intuition, and test injections.Computers are more limited.
7
Matching Peaks for Complex Work
Each component in a separation must be tracked
Manual peak tracking can be tedious in CA MethDev; it’s impossible in
AMD!min6 8 10 12 14 16 18 20
mAU
0
25
50
75
100
125
150
175
min6 8 10 12 14 16 18 20
mAU
0
25
50
75
100
125
150
175
YMC Pro C18 pH 2.5
XTerra RP18 C18 pH 7
8
MS-Mutual Automated Peak Matching
MS offers significant advantages over UV:Greater resolutionS/NOrthogonal signal strength
The MS-MAP algorithm was designed to take advantage of the differences in these detectors
9
Preparing for Peak Matching -IntelliXtract
10
IntelliXtract Key Features
Built on CODA peak extraction technology, but takes it to the next level Uses a mixture of
Chemometrics (CODA, Mass Cluster Analysis, Isotopic Matching…), andMass Spectral knowledge
Follows the same logic that an expert would use for manual data interpretationA self-optimizing algorithm
Software parameters are calculated dynamically
11
DefinitionsComponent
A component is defined as a single chemical entity eluting in chromatographic space
5.04.54.03.53.02.52.01.51.00.5Retention Time (min)
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
Rel
ativ
e In
tens
ity
TIC
Extracted component
[M+H]+
12
IntelliXtract
Examine each mass-peak individuallyValidate chromatographic character of peakDesignate each peak’s “role”:
(M+H)+
AdductDimer/trimerProbable fragment ion (PFI)Isotopes
13
MS-MAP: Reconciling Components
Perfect co-elution is a problem for IX (in a single experiment)Sometimes, M+H+ assignment can be wrongSolution:
Full spectral matching post-IXLogic-based reconciliation (if new component arises in run2, then re-examine run1’s PFIs)Orthogonal detection (UV-MAP)
14
Reconciliation Logic
1. From IX:Masses for each tR and each expt. with (M+H)+
label2. Match all “complete” (M+H)+s spectra3. Search all missing (M+H)+ masses for PFI or
adduct assigned peaks in other peaks4. Transfer peak information where masses are
found5. Label redundant masses6. Reconcile with LC/UV data
15
Illustrating MS-MAP
16
NONAME00
Using a second experiment to supplement IX
Comp. A
Comp. B
288289
287
365365366366
320321
Component A and B coelute perfectly in expt. 1. IX will assign the component A masses “probable fragments”.
17
NONAME01
NONAME00
Using a second experiment to supplement IX
Comp. A
Comp. B
288289
287
365365366366
320321
The emergence of the new pattern enables us to track the component from expt2 back to expt 1.
18
NONAME19 1
NONAME17
NDMC “Virtual Resolution”
Comp. A
Comp. C
Comp. B
288289
287
365365366366
320321
303
305305304
• Component A has chromatographic resolution in experiment 1.• Component B has CR in exp 2.• Component C is resolved only through the use of both expts!
19
Virtual Resolution
A component needs be resolved from its partner in only one experiment to be tracked, and need never be resolved from all other components in even one single run.
20
GC x GC of yeast cell extract
21
NDMC Projections – Orthogonal Methods with Peak Tracking
Collect orthogonal dataMatch peaksEstablish peak table:
Component, tR, Width, etc.Widths and tRs in any two experiments can be combined
22
2 Dimensional MAP Chromatography
NONAME19 1
Comp. A
Comp. C
Comp. B
288289
287
365365366366
320321
303
305305304
NONAME17
NONAME17
23
2D MAP Chromatography
Coelution in method 2
Coelution in method 1
24
NONAME19 1
NONAME17
NONAME17
NONAME17
NONAME17
NONAME17
NONAME17
…more experiments
NONAME17
Comp. A
Comp. C
Comp. B
288289
287
365365366366
320321
303
305305304
25
NONAME19 1
NONAME17
NONAME17
NONAME17
NONAME17
NONAME17
NONAME17
…more experiments
NONAME17
Only in the 8th
experiment were comps A and C
resolved.
26
NONAME19 1
NONAME17
NONAME17
NONAME17
NONAME17
NONAME17
NONAME17
…more experiments
NONAME17
In this case, 8 experiments were required in order
to resolve component C!
27
Resolving Power of Multidimensional Chromatography
28
Adding Dimensions
Any peak-matched experiments can be used to contribute to the component tableResolution of a given pair of components is necessary in only one experimentThe more (orthogonal) experiments, the greater the chance of virtual resolution
29
Theoretical Peaks
~100 peaks in 1D HPLC?2D MAP HPLC: 10000 peaks32D: 1064 peaks
)ln(4
1M
RC t
tNn +=
Three very important assumptions are made:• Signals are independent of coelution• Retention times don’t correlate• No isobaric compounds…therefore, in practice, far fewer peaks will be resolved.
30
How best to design the chromatographic conditions?
Ideally, the chemical environment is stable, but the chromatographic environment changes dramatically.Logically, the column is the thing to change.
31
Chromatographic Conditions and Peak Matching
MAP works best with smallsolvent condition changes:
ColumnColumnSolvent strengthSolvent strengthTemperatureTemperatureSolventSolventpHpH
32
Some Early NDMC ResultsMethod Development Suite LC/MS 10.0
2-Dimensional MAP Chromatography:Acetic acid gradient methanol vs. acetonitrileGeneric gradients; no optimization of chromatography or detector
10 compound set:m-toluamideUmbelliferonePropazineMirtazapineDexamethasone4-aminobenzoic acidBenzoylecognineLinomycinPrilosec4-hydroxy-3-methoxyphenylacetic acid
33
Some Early NDMC Results (1)
2-Dimensional MAP Chromatography:Waters Alliance 2690 with ZQ Single Quad MS Acetic acid gradient methanol vs. acetonitrileGeneric gradients; no optimization of chromatography or detector
10 compound set:m-toluamideUmbelliferonePropazineMirtazapineDexamethasone4-aminobenzoic acidBenzoylecognineLinomycinPrilosec4-hydroxy-3-methoxyphenylacetic acid
Failed to ionize
Correctly identified (M+H)+
34
Some Early NDMC Results (2)
6-Dimensional MAP Chromatography:
Waters Alliance 2690 with ZQ Single Quad MS Supelco C18/Amide/Phenyl Columns:
5 cm x 4.6 mm ID 5 micron0.1% Acetic acid, ammonium acetate, 5 min acetonitrile gradient 5%-95%Generic gradients; no optimization of chromatography
Sample set:59 non-isobaric compounds: 10 ug/mL Commercially available107.07 < MW < 853.33All known to give signal when resolved under these detector conditions
35
Standards Set Results2DMC (average)
Of 59 compounds:33 found with correct MW, tRs23 false positives26 unmatched
36
Standards Set Results3DMC (average)
Column was the variable changedBuffer was held constantOf 59 compounds:
37 found with correct MW, tRs6 false positives22 unmatched
37
Standards Set Results6DMC
Column and mobile phase changeOf 59 compounds:
42 found with correct MW, tRs5 false positives17 unmatched
Note:Total Peak width = 13.7 minutesEffective Chr’gram ~ 5.5 minutesOn average, almost 3 components coeluteat any point in the ch’gram.
38 tR amide/acetic acid
tRC18/ acetic
acid
MAP Projection6-DMC
39
MAP Projection6-DMC
tR C18/ammonium acetate
tRAmide/
acetic acid
40
The value of added dimensions
A few extra components were locatedMany false positives were excluded
41
Tannins Project
Ontario spruce tannins:High MW fraction was collected
Instrument:Agilent 1100, series pump autosampler, pump (quat), degasser.Detectors:
Bruker DADHP/Bruker Esquire LC 3D ion trap MS (negative mode)Additional acetonitrile ~20uL/min was added by a Bruker NMR-MS interface (BNMI) just prior to MS to aid electrospray.
42
Chromatographic Conditions
Gradient:1) Isocratic 90:10 (Water:acn) for 5 mins2) Linear Gradient 5-60mins (Water : acn) to 10:903) Linear Gradient 60-80mins (Water : acn) to 0:1004) Isocratic 0:100 (Water:acn) from 80-100mins
4-dimensional MAP Chromatography:Xterra PhenylAscentis AmideSupelco PFPPAscentis C18
43
Ascentis RP-Amide
Supelco-PFPP
888072645648403224168Retention Time (min)
Ascentis C18
Xterra Phenyl
The TICs
44
96 unique mass components found38 isobaric sets
NDMC Projection
tRC18
tR Amide
45
How Many Dimensions?
Large numbers of sets of somewhat orthogonal HPLC conditions can be designedThe resolving power of NDMC is limited by the number of sets of orthogonal conditions that still allow for consistent spectrapH and column are primary dramatic selectivity modifiersLots of stationary phases exist, but eventually, they become redundantSpectral matching with pH changes will likely be challenging to perfectIsobaric compounds will still be a challenge in LC
46
Where can NDMC be useful?
COMPLEX SAMPLESInitial investigation of (unknown composition) samplesAs a fast detection technique for preliminary detection of componentsAs part of initial screening of the sample for detection of impuritiesIn supplement to manual LCMS analysis
47
Some Remaining Questions
Where is the best compromise between run time and added dimensions?
Suppression of ionizationEfficiency of peak matching
How can we further reduce false positives?How to reduce missed components?
composite samples…it is possible to combine multiple detector settings automatically…
48
Conclusion
Orthogonal chromatographic techniques combined with MAP:
Method design toolAutomated detection device
A useful tool for complex sample investigation with no prior knowledge of componentsRequires no specialized hardwareRidiculous theoretical resolving power is practically limited
49
Acknowledgements
ACD/Labs:A. Bogomolov and A. Galin
Supelco:David BellJacinth McKenzie
University of Toronto:Andre Simpson
University of WashingtonRob Synovec