Lipid spectral prediction performance

Figure 1a
Figure 1b
Figure 1c

Figure 1. Head-to-tail plot of experimental and predicted electrospray ionization-tandem mass spectroscopy (ESI-MS/MS) spectra of PC(16:0/16:0). (a) Head-to-tail plot showing an experimental ESI-MS/MS spectrum of dipalmitoyl phosphatidylcholine (PC(16:0/16:0)) measured at 40 eV, and the matching ESI-MS/MS spectrum predicted by CFM-ID 2.0. The computed spectral similarity score is 0.07. (b) Head-to-tail plot showing an experimental of ESI-MS/MS spectrum of dipalmitoyl phosphatidylcholine measured in positive ion mode ([M+H]+) at 40 eV, and the matching ESI-MS/MS spectrum predicted by CFM-ID 3.0. The computed spectral similarity score is 0.88. (c) Head-to-tail plot showing an experimental of ESI-MS/MS spectrum of dipalmitoyl phosphatidylcholine measured in positive ion mode ([M+H]+) at 40 eV, and the matching ESI-MS/MS spectrum predicted by LipidBlast. The computed spectral similarity score is 0.13.


Figure 1a
Figure 1b
Figure 1c

Figure 2. Head-to-tail plot of experimental and predicted ESI-MS/MS spectra of PS(16:0/18:1(9Z)). (a) Head-to-tail plot showing an experimental of ESI-MS/MS spectrum of 1-palmitoyl-2-oleoylsn- glycero-3-phospho-L-serine PS(16:0/18:1(9Z)) measured at 40 eV, and the matching ESI-MS/MS spectrum predicted by CFM-ID 2.0. The computed spectral similarity score is 0.10. (b) Head-to-tail plot showing an experimental ESI-MS/MS spectrum of 1-palmitoyl-2-oleoyl-sn- glycero-3-phospho-L-serine PS(16:0/18:1(9Z)) measured at 40 eV, and the matching ESI-MS/MS spectrum predicted by CFM-ID 3.0. The computed similarity score is 0.92. (c) Head-to-tail plot showing an experimental ESI-MS/MS spectrum of 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine PS(16:0/18:1(9Z)) measured at 40 eV, and the matching ESI-MS/MS spectrum predicted by LipidBlast. The computed similarity score is 0.91.


Table 1. Computed spectral similarity scores between experimental and predicted ESI-MS/MS spectra at three energy levels (10, 20, and 40 eV). The results show higher similarities, and thus an improvement when using a rule-based approach (CFM-ID 3.0) over a combinatorial one (CFM-ID 2.0) for the prediction of lipid ESI-MS/MS spectra. The spectral similarities of the LipidBlast-generated consensus spectra further illustrate this trend. When available, the same LipidBlast-generated consensus spectrum was used for comparisons at each energy level. N/A corresponds to cases where (1) the adduct type was not covered by CFM-ID 2.0 at all, or (2) the adduct type was not covered by LipidBlast for the chemical class to which the test compound belongs.

CompoundAdductEnergy (eV)CFM-ID 3.0 (Score)CFM-ID 2.0 (Score)LipidBlast (Score)
PA(16:0/18:1(9Z))[M-H]-101.000.360.00
PS(16:0/18:1(9Z))[M-H]-101.000.310.00
CL(18:0/18:0/18:0/18:0)[M-2H](2H)100.98N/A0.00
DG(18:0/20:4/0:0)[M+Na]+100.920.00N/A
PA(16:0/18:1(9Z))[M-H]-200.550.020.00
PS(16:0/18:1(9Z))[M-H]-200.980.030.00
CL(18:0/18:0/18:0/18:0)[M-2H](2H)200.97N/A0.12
DG(18:0/20:4/0:0)[M+Na]+200.930.00N/A
PA(16:0/18:1(9Z))[M-H]-400.960.030.90
PS(16:0/18:1(9Z))[M-H]-400.920.100.91
CL(18:0/18:0/18:0/18:0)[M-2H](2H)400.91N/A0.89
DG(18:0/20:4/0:0)[M+Na]+400.180.00N/A
PC(16:0/16:0)[M+H]+400.880.070.13
TG(18:1/18:1/18:2)[M+NH4]+400.780.010.84

Compound identification performance

Table 2. Comparison of CFM-ID 3.0, CFM-ID 2.0, and MS-FINDER scoring functions upon identification of 185 compounds from 208 ESI-MS/MS spectra. Reported are the total number of challenges in which the corresponding implementation of the scoring function ranked the query compound in the top 1, top 3, and top 10. The average and median ranks for the query compound are also reported. A chemical classification is assessed as correct if the predicted category matches a category originally assigned by ClassyFire. N/A, not applicable; * performance when applied over the expanded spectral library database including the 208 experimental ESI-MS/MS from the CASMI 2016 contest (category 3).

Version# Top 1# Top 3# Top 10Average RankMedian Rank# Correct Classifications
CFM-ID 3.01491942041.81168
CFM-ID 2.0-20191231712012.41N/A
CFM-ID 2.0-201612016018213.641N/A
MS-FINDER1461621746.41N/A
CFM-ID 3.0 *20820820811N/A
CFM-ID 2.0-2019 *20820820811N/A