To the Editor-in-Chief Sir, An improved two-step calibra- tion method for matrix-assisted laser desorption/ionization time- of-flight mass spectra for proteo- mics Since its introduction in 1993, 1 matrix- assisted laser desorption/ionization (MALDI) peptide mass fingerprinting (PMF) has become one of the standard methods for proteome research. Among other factors, such as the size of the protein and sequence database, the reliability of a protein identifica- tion using PMF primarily depends on the number of peptide ion signals matching to the respective protein hit and the accuracy with which the m/z values of these signals can be deter- mined. In an ideal, continuous extraction linear time-of-flight instrument, a sim- ple internal two-point calibration should yield good mass accuracy over the entire m/z range, since m/z is proportional to t 2 (t is flight time). In practice, however, good mass accu- racy can only be obtained with instru- ments equipped with time-delayed ion extraction, low mass deflection pulsers and single- or dual-stage reflectrons that partly compensate for the space and energy distributions of ions generated by MALDI. These instruments have attained resolving power beyond 10000 FWHM and sensitivities in the attomole range, but delayed ion extraction, low mass ion deflection, and the reflectron directly influence individual molecu- lar ion velocities and, thus, complicate the correlation between m/z and flight time. Therefore, mass accuracies in mass spectra of peptides have gener- ally not reached beyond 20 ppm. Recently, some effort has been put into optimizing calibration proce- dures for increasing mass accuracy over a broad mass range. 2±5 For example, simplex optimization and quadratic equations were used to describe the correlation between sys- tematic mass errors and m/z. 4,5 How- ever, mass accuracies in the low ppm range could not be achieved. Recently, Gobom and co-workers 3 suggested a procedure using polyethylene glycol (PPG) to determine systematic mass errors and eliminate those by a 15th order correction function. Although initial velocities of PPG and peptides might be different, good mass accura- cies were achieved when this method was applied to PMF spectra. How- ever, PPG yields only one signal per mass increment of 58 Da. In order to achieve a sufficiently large data set for calculation of higher-order fitting functions, a method allowing the implementation of more data points generated from compounds with a similar chemical nature to analytes should be beneficial. Here we report a calibrationmethodthatisbasedonthe empirical determination of systematic m/z-dependent mass errors in intern- ally calibrated MALDI PMF spectra and the elimination of these errors using a higher-order m/z correction function. This method was found to greatly improve mass accuracy in MALDI peptide mass fingerprints and, thus, enables more reliable pro- tein identification. In order to determine systematic mass errors corresponding to distinct m/z values in MALDI-TOF spectra, a training set of peptide mass finger- printspectrafrom50randomlypicked in-gel digested (trypsin) protein sam- ples was acquired. In addition, 25 independent measurements of 5 fmol of a solution digest of myosin with trypsin were performed. Overall, the training set consisted of more than 1800 peptide mass measurements. All spectra were acquired using a Reflex IV 2 MALDI-TOF instrument (Bruker Daltonik, Bremen, Germany) equipped with a gated detector. The instrument was run in reflector mode at a full reflector voltage of 27.5 kV. The high voltages of the instrument were always switched on between analyses in order to ensure stable instrument voltages. When sample plates were exchanged, high voltages were switched on at least 30 min prior to spectra acquisition. Samples were spotted onto 600 mm anchor targets (Scout 384-MTP An- chorChip; Bruker Daltonik, Bremen, Germany) according to a dried-drop- let protocol using a-cyano-4-hydroxy- cinnamic acid (CHCA) as matrix. Three different target plates were used and sample positions were evenly distributed over the target plates in order to eliminate position and plate dependent parameters. It should be noted that the outer two rows and columns of the target plates were not used because we observed far worse mass accuracies and sensi- tivities at these positions than at the other positions. All spectra were re- corded in automatic mode. Peak label- ing and internal two-point calibration on trypsin autolysis products (m/z 842.5100 and 2211.1046) were per- formed without user interference using the SNAP algorithm and Aura scripts implemented into the XTOF software. Database searches were car- ried out using Profound (Genomic Solutions, Ann Arbor, USA) with mass tolerance set to 70 ppm. Observed masses were compared with calculated masses of matching peptides of the identified proteins and relative mass errors (in ppm) were calculated for each peptide ion signal. Systematic mass errors corresponding to distinct m/z values were determined by averaging the measured mass errors of individual peptides corre- sponding to the same m/z value. Plotting of the relative mass error over m/z indicated a continuous distribu- tion of data points (Fig. 1(a)) that was subsequently subjected to a polyno- mial fitting procedure using standard Excel worksheet extensions. A 7th order polynomial function was found to approximate the correlation be- tween m/z and relative mass errors (Fig. 1(a)). Application of this poly- nomial function to the training set drastically reduced relative mass er- rors. All data points were now found within 20 ppm (Fig. 1(b)). Using internal calibration only, the average relative mass error of the training set Copyright # 2002 John Wiley & Sons, Ltd. RAPID COMMUNICATIONS IN MASS SPECTROMETRY Rapid Commun. Mass Spectrom. 2002; 16: 1892±1895 Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/rcm.798 RCM Letter to the Editor