Peak detection routines

A typical MALDI spectrum of a bacterial sample has a number of peaks that vary greatly in intensity superimposed on a relatively noisy baseline. This can be problematic for many peak detection routines. Therefore methods that eliminate the need for peak detection also eliminate problems associated with poor peak detection performance. Full-spectrum identification algorithms use the (usually smoothed) spectral data without first performing peak detection. 

However, they do generally require an effective peak detection routine to ensure that both large and small key peaks are used, and that noise spikes are not detected as peaks. The reader is referred to Jarman et al.16 for a discussion and comparison of some different peak detection routines applied to MAID I data of bacterial samples. 

In addition to commercial software products we also took advantage of custom designed software (MicrobeMS Lasch 2015) for the evaluation of our microbial mass spectra. MicrobeMS is Matlab-based and involves a specifically optimized peak detection routine. One of the key features of peak detection in MicrobeMS is a sigmoid intensity threshold function which was introduced to model the m/z dependence of the analytical sensitivity of MALDI-TOF MS. This threshold function defines intensity thresholds at each m/z value. In the MicrobeMS implementation, an intensity threshold at low m/z values is larger than at high m/z values. Another feature of the MicrobeMS peak detection routine allows to precisely define the number of resulting peaks per spectrum. This particular feature makes peak detection partially independent from the SNR which turned out to be extremely useful for subsequent classification analysis. 

Fraction collection can either be on the basis of time or volume, or can utilize a peak detection routine, most commonly based on the UV signal, to trigger the start and finish of each fraction. Although sophisticated peak detection routines are utilized on analytical systems, they are not always applicable to preparative or production systems. Peak shape, imder analytical conditions, is usually well defined over a short time period, in preparative applications this is often not the case. The simplest and most reliable method is timed collection and can be used for reproducible separations or where several fractions are to be taken and pooled later. 

Compared to flame excitation, random fluctuations in the intensity of emitted radiation from samples excited by arc and spark discharges are considerable. For this reason instantaneous measurements are not sufficiently reliable for analytical purposes and it is necessary to measure integrated intensities over periods of up to several minutes. Modern instruments will be computer controlled and fitted with VDUs. Computer-based data handling will enable qualitative analysis by sequential examination of the spectrum for elemental lines. Peak integration may be used for quantitative analysis and peak overlay routines for comparisons with standard spectra, detection of interferences and their correction (Figure 8.4). Alternatively an instrument fitted with a poly-chromator and which has a number of fixed channels (ca. 30) enables simultaneous measurements to be made. Such instruments are called direct reading spectrometers. 

Most integrators are developed for use in GC and have sophisticated routines for peak detection and for baseline corrections if there are negative slopes. They are able to detect small peaks even in the descending part of a solvent peak. How-... 

The Diamat system (Bio-Rad Laboratories, Hercules, CA) is also suitable for hemoglobinopathy screening. Although it was originally designed mainly for determinations of HbAl. it can be used for detection of some Hb variants. It has been shown to separate seven variants, but many overlap partially with the HbAi peak [17]. Aberrant peaks detected with the Diamat method during routine assays for HbAic led us to the detection of two Hb variants, Hb Broussais and Hb Cemenelum [18]. With PolyCAT A, both variants could be totally separated (Fig. 3a and b). The elution pattern can be used for preliminary identification of these Hb variants, especially if standards are available. 

LC is frequently coupled with UV, diode arrays, ELSD, MS, and more rarely with NMR for the peak detection. The tropane nucleus itself has no chromophore and the sensitivity of UV detection and diode array detectors (DAD) depends on the esterified moiety. UV and DAD are suitable for the detection of alkaloids with chromophore-bearing moieties such as hyoscyamine and scopolamine. For such alkaloids, routine analytical work and quantitation are mostly performed between 210 nm and 220 nm [66-68,71], but an analysis at a wavelength of 254 nm was also reported [70]. Hosseini et al. [72] reported limit of detection (LOD) and limit of quantification (LOQ) values of 5.15 and 17.4 ppm for atropine and 1.92 and 6.4 ppm for scopolamine, respectively, for a vahdated HPLC method with UV detection at 215 nm, which is comparable with previous reports [15]. 

There are subtle variations of this type of detection system, but its major benefit is that it requires only one scan to determine both high and low concentrations. It, therefore, not only offers the potential to improve sample throughput, but also means that the maximum data can be collected on a transient signal that only lasts a few seconds. This is described in greater detail in Chapter 12, where we discuss different measurement protocols and peak integration routines. 

A continuous or transient signal The temporal length of the sampling event Volume of sample available Number of samples being analyzed Number of replicates per sample Number of elements being determined Detection limits required Precision/accuracy expected Dynamic range needed Integration time used Peak quantitation routines... 

With no stable isotope pair within the U system or a suitable AME, a standard-sample bracketing protocol is usually employed to correct for mass bias. Human urine generally contains very low concentrations of U (generally 1-5 ng/L), so an isotope dilution strategy is required, together with ion-counting detection (ideally a Daly photomultiplier or discrete dynode secondary electron multiplier) and a multi-static (rather than multi-dynamic) peak-jumping routine, for precise measurement of the total U concentration and the minor isotopes of and even... 

The special problems for vaUdation presented by chiral separations can be even more burdensome for gc because most methods of detection (eg, flame ionization detection or electron capture detection) in gc destroy the sample. Even when nondestmctive detection (eg, thermal conductivity) is used, individual peak collection is generally more difficult than in Ic or tic. Thus, off-line chiroptical analysis is not usually an option. Eortunately, gc can be readily coupled to a mass spectrometer and is routinely used to vaUdate a chiral separation. 

---