For fentanyls, 4 out of 16 compounds of the mixture were correctly identified by library matching, namely fentanyl (identified as “Innovar” which is a trade name of a drug containing fentanyl) remifentanyl, sufentanyl, and alfentanyl while no hits were found for the remaining drugs. Among the drugs identified by the library matching, only fentanyl (m/z 337.227) was included in a network (Figure 2). The original network contained 25 nodes, however, as reported above for synthetic cannabinoids, nodes with lower intensity and nodes that were likely to belong to adducts and in-source fragments were deleted; the cleaned network included 16 nodes.

Ten of the standards included in the mixture were found in this network as reported in Table 2. These drugs correspond to the yellow colored nodes, while nodes from group 2 are colored in blue and group 3 in red. From a visual analysis it can be noticed that nodes belonging to this network are from the three groups, suggesting that seizures A (group 2) and B (group 3) contained fentanyl derivatives. In the case of fentanyls, differently from synthetic cannabinoids not all the standards of the mixture were included in the network. This can be explained observing the molecular structures of the identified fentanyls (Figure 3) which lead to completely different fragmentation spectra. Fentanyls not included in the network were those not having the typical phenylethyl piperidine moiety which gives rise to MS2 spectra with the characteristic m/z 188.143 and 105.070 fragments.

This results in showed a limitation of MN for the identification of fentanyl derivatives with a different base structure. Future algorithms may be able to find additional connections, even when the overall fragmentation behavior is altered, as long as some overlap in fragmentation is present.

However, good data were obtained for the seizures; some unknown hits were connected to the standard fentanyl network suggesting that they were structurally related to the standards, with a cosine score >0.9. A number of nodes were exclusively found in the seizures while some nodes were common to seizures and standard mixture as listed in Table 3.

In addition to the identification of unexpected drugs, an important feature of MN analysis is that the structures of the unknowns may be hypothesized, based on the precursor mass difference between nodes. The node with mass 299.192 corresponded to despropionyl para-fluorofentanyl, which is known to be a metabolite of 4-fluoroisobutyryl fentanyl, but also a precursor in the synthesis of para-fluorofentanyl; the presence of this compound in both seizures suggested that they contained a para-fluorofentanyl derivative. From the analysis of the network, it can be noticed that despropionyl para-fluorofentanyl is connected to the node with m/z 393.197 (cos 0.95) which was found in both seizures and was among the most intense peaks in seizure B. The delta mass between these two nodes was 94.005, which correspond to a C₅H₂O₂ moiety; the node with m/z 393.197 was also connected to furanylfentanyl node with a cosine score of 0.93 and, in this case the delta mass between these two nodes was 17.991 which possibly arose from the addition of a fluorine and the loss of a proton. This observation suggested that peak with m/z 393.197 is a fluoro-furanylfentanyl; the presence of despropionyl para-fluorofentanyl in the seizures indicate that it could be para-fluorofuranylfentanyl, arising from the addition of a furanyl moiety (C₅H₃O₂) to the precursor despropionyl para-fluorofentanyl.

Another interesting observation was that two different nodes had m/z 351.243. One of these corresponded to α-methylfentanyl which was among the selected standards. In fact, it was found in group 1, the corresponding spectrum showed the characteristic fragment 202.159 (Figure 4) which arose from the loss of the methylated phenylethyl piperidine moiety. The spectrum of the node which shared the same mass of α-methylfentanyl is different and only the typical m/z 188 and 105 fragments are present, suggesting that the phenylethyl piperidine moiety was unmodified; this compound, which was found in both seizures and mainly in seizure A, can be putatively annotated as butyrylfentanyl or isobutyrylfentanyl.

Seizure A also contained 4-ANPP which is a well-known fentanyl precursor (Drug Enforcement Administration, 2010) and 3 unknown compounds, namely m/z 437.280, 619.365, and 403.238 whose spectra are shown in Figure 5. All three spectra are compatible with fentanyl derivatives on the basis of the presence of the characteristic m/z 105 and 188 fragments, but the precursor masses do not match with any known fentanyl. Compared with the peaks putatively annotated as butyrylfentanyl and 4-Fluoro-furanylfentanyl, their intensity is 2–3 order of magnitude lower. Interestingly the node with mass 403.238 is central in the fentanyl network, supporting the thesis that it may be a fentanyl derivative. The node is linked to 4-ANPP (Δmass 122.037, C₇H₆O₂), furanylfentanyl (Δmass 28.032, C₂H₄) and butyrylfentanyl (Δmass 51.995, C₃O) among others, with cosine score >0.9. By taking into account the delta masses between nodes (Figure 2) and the fragmentation spectrum a possible assignation could be dimethylfuranyl-fentanyl.

A hypothesis is that the three unknown compounds in seizure A are synthetic byproducts and even if it was not possible to assign a possible structure for all these compounds at this stage, the detection of possible impurities is a tool for investigations since they might be an important marker for its source and are likely to be specific for a particular synthesis site or wholesaler.