Componentization-Method Optimization. Parameter. Sample colchicine hydrocodone ricinine yohimbine. No peak limit, default settings. Standard Mixture, 2.
Toward Characterizing Unknowns in Food Matrices: Software Strategies for LC/MS Data Ann M. Knolhoff, John H. Callahan, Timothy R. Croley Food and Drug Administration, Center for Food Safety and Applied Nutrition Jerry Zweigenbaum, Agilent Technologies
Non-Targeted Analysis • Ultimate goal: chemical Componentization screening method with automated data processing Interpretation of Detected Ions • LC/MS – Complex sample matrices – Large data sets
• How to determine compounds of interest?
Formula Generation
Database Search
Experimental Workflow Main Goal: Determine differences between control and suspect sets of samples
ACN extraction, centrifuge, filter 5 replicates for each lot
3
7
Standard mixture 10, 100, 500 ppb Analyte Formula colchicine C22H25NO6 hydrocodone C18H21NO3 ricinine C8H8N2O2 yohimbine C21H26N2O3
Blanks Standard Mixture Samples
LC/MS
Data Analysis Workflow Componentization
MassHunter Interpretation of Detected Ions
Formula Generation
Database Search
Mass Profiler Professional (MPP)
Parameter No peak limit, default settings No peak limit, 50 quality score No peak limit, 30 quality score No peak limit, 0 quality score Removal of blank compounds RT restriction: 2-29 min RT restriction: 4-20 min
Sample Standard Mixture, 2 Formula, 10 ppb, 1 Standard Mixture, 2 Formula, 10 ppb, 1 Standard Mixture, 2 Formula, 10 ppb, 1 Standard Mixture, 2 Formula, 10 ppb, 1 Standard Mixture, 2 Formula, 10 ppb, 1 Standard Mixture, 2 Formula, 10 ppb, 1 Standard Mixture, 2 Formula, 10 ppb, 1
colchicine hydrocodone ricinine yohimbine
Componentization-Method Optimization
Sample Complexity Infant Formula >80 features in all 15 samples from 3 lots
Orange Juice >1900 features in all 35 samples from 7 lots
MPP General Workflow 2 out of # samples have acceptable values
Retain entities that appear in at least 66% of samples in at least one condition
Pick p-value where 0 false positives are expected
Fold change of 2 in suspect set
MPP General Workflow 2 out of # samples have acceptable values
Retain entities that appear in at least 66% of samples in at least one condition
Pick p-value where 0 false positives are expected
Fold change of 2 in suspect set
MPP General Workflow 2 out of # samples have acceptable values
Retain entities that appear in at least 66% of samples in at least one condition
Pick p-value where 0 false positives are expected
Fold change of 2 in suspect set
Infant Formula: Spiked Formula vs Lots 2 & 3 • 10 compounds found – Includes all compounds in standard mixture Spiked formula
2/4 in 50 ppb standard
Spiked formula
1/4 in blanks
1. Recursion 2. Find by Formula 3. Reimport into MPP
Formula Generation and Database Searching 10 compounds found Formula
Mass Score Difference (ppm)
# > 90 Score
# > 80 Score
C19H3N3O4
86.18
-4.11
0
2
C8H8N2O2
96.57
-2.9
1
1
C18H24N2O3
70.88
3.37
0
0
C18H21NO3
97.27
-2.71
1
2
C21H26N2O3
97.53
-2.5
1
2
C21H23NO6
98.84
-0.55
3
7
C21H23NO6
98.54
-0.21
2
7
C23H21N5O2
99.26
0.78
3
6
C20H18N8
96.67
0.21
2
4
C12H10N4O2
97.24
-0.43
1
3
Correct formula was the second hit, score=97.36
Formula Generation and Database Searching 10 compounds found Formula
Mass Score Difference (ppm)
Database Formula Search, # of hits Compound (in mix) ChemSpider PubChem
Metlin
C8H8N2O2
595
8
1
C18H21NO3
0
19
13
C19H3N3O4
86.18
-4.11
C8H8N2O2
96.57
-2.9
C18H24N2O3
70.88
3.37
C21H26N2O3
5230
6
9
C18H21NO3
97.27
-2.71
C22H25NO6
0
3
3
C21H26N2O3
97.53
-2.5
Other compounds
C21H23NO6
98.84
-0.55
C19H3N3O4
0
0
0
C21H23NO6
98.54
-0.21
2941
2
2
C23H21N5O2
99.26
0.78
C18H24N2O3
C20H18N8
96.67
0.21
C21H23NO6
1748
0
8
C12H10N4O2
97.24
-0.43
C20H18N8
21
0
0
C23H21N5O2
856
3
0
Correct formula was the second hit, score=97.36
Formula Generation and Database Searching 10 compounds found Formula
Mass Score Difference (ppm)
Database Formula Search, # of hits Compound (in mix) ChemSpider PubChem
Metlin
C8H8N2O2
595
8
1
C18H21NO3
0
19
13
C19H3N3O4
86.18
-4.11
C8H8N2O2
96.57
-2.9
C18H24N2O3
70.88
3.37
C21H26N2O3
5230
6
9
C18H21NO3
97.27
-2.71
C22H25NO6
0
3
3
C21H26N2O3
97.53
-2.5
Other compounds
C21H23NO6
98.84
-0.55
C19H3N3O4
0
0
0
C21H23NO6
98.54
-0.21
2941
2
2
C23H21N5O2
99.26
0.78
C18H24N2O3
C20H18N8
96.67
0.21
C21H23NO6
1748
0
8
C12H10N4O2
97.24
-0.43
C20H18N8
21
0
0
C23H21N5O2
856
3
0
Correct formula was the second hit, score=97.36
Metlin hit: desmethylcolchicine and gloriosine, metabolites of colchicine
Infant Formula 10 ppb Spiked Formula vs Lots 2 & 3 • 11 compounds found – Includes all compounds in standard mixture – No overlap with blank
Spiked Formula
2/4 in 50 ppb standard
Orange Juice Analysis Retain entities that appear in at least X% of samples in at least one condition OJ vs all spikes • 66%: 38 compounds • 80%: 20 compounds • All compounds in mixture detected OJ vs 10 ppb spikes • 66%: 8 compounds • All compounds in mixture detected OJ lot 6 vs other lots • 66%: 1 compound
Spiked OJ
Lot 7
Orange Juice: Lot Comparison
Orange Juice: Post-Recursion Spiked OJ
OJ, 2/5 within lot
Orange Juice: Post-Recursion Spiked OJ
OJ, 2/5 within lot
# > 90 # > 80 Score Score
Formula
Score
Mass
1
C11H10O4
96.88
206.0584
2
3
2
C8H8N2O2
98.09
164.0590
1
1
3
C18H21NO3
99.25
299.1525
1
2
4
C21H26N2O3
97.58
354.1951
2
2
5
C21H23NO6
99.41
385.1526
4
8
6
C21H23NO6
97.65
385.1525
2
4
7
C22H25NO6
97.85
399.1690
4
6
0
7
8 9
973.3539 C12H16N12S3
86.86
No database hits
424.0783
No formula generated
Database of Compounds in Commodity 2 out of # samples have acceptable values
Retain entities that appear in at least 50% of samples in at least one condition
Removal of entities from blank and standard (2/total)
Recursion, Find by Formula, Reimport data into MPP
Database of Compounds in Commodity 2 out of # samples have acceptable values
Retain entities that appear in at least 50% of samples in at least one condition
Removal of entities from blank and standard (2/total)
Recursion, Find by Formula, Reimport data into MPP
Database of Compounds in Commodity 2 out of # samples have acceptable values
Retain entities that appear in at least 50% of samples in at least one condition
Removal of entities from blank and standard (2/total)
Recursion, Find by Formula, Reimport data into MPP
Combined Entity List
Peak extraction Optimize method Compounds in QC detected
N
Y MPP More processing needed # features < 20
N
Y Recursion, Find compounds, MPP for confirmation & commodity search
Formula generation
MS/MS
Database searching
Identification
Conclusions • Workflow design – QC standard and blanks – Sufficient replicates, representative samples • Samples were within same brand/type
• Considerations – Complexity of sample matrix – Limit number of targets as much as possible without filtering out statistically relevant compounds
• Statistical workflow can help with determining molecules of interest, especially with large data sets
