This thesis describes the development of novel strategies for the analysis of peptides by MALDI mass spectrometry. The developed techniques are applied to the identification of protein and proteomic biomarkers for melanoma. A commercial atmospheric pressure (APMALDI) source (MassTechnologies, Burtonsville, MD, USA) was modified to allow operation with a high powered nitrogen laser and independent PC control of the sample stage. A software interface was developed using LabVIEW 6.1 that allows full control of the target position with respect to the laser fibre optic interface, allowing the target to be adjusted within any point within a particular sample spot to enhance signal quality. The modified AP-MALDI-QIT interface was evaluated for the analysis of standard peptide mixtures and tryptic digests of proteins. AP-MALDI-QIT analysis of tryptic peptides following capillary liquid chromatographic (LC) separation and direct analysis of a protein digest is reported. Peptide fragments were identified by peptide mass fingerprinting from mass spectrometric data and sequence analysis obtained by tandem mass spectrometry of the principal mass spectral peaks using a data-dependent scanning protocol. These data were compared with those from mass spectrometric analysis using capillary LC/MALDI-time-of-flight (TOF) and capillary LC/electrospray ionisation (ESI)-quadrupole TOF. For all three configurations the resulting data were searched against the MSDB database, using MASCOT and the sequence coverage compared for each technique. Complementary data were obtained using the three techniques. A bottom-up proteomic methodology for the peptide profiling of human serum samples using MALDI mass spectrometry was developed. Reproducibility studies were carried out to define the MALDI measurement precision. Pre-analytical sample handling factors, such as room temperature incubation and freeze thaw cycles have also been investigated. The methodology developed was applied to the analysis of serum peptides from stage IV melanoma patients and healthy control subjects. Prediction of human melanoma metastatic cancer from peptide profiling using artificial neural networks (ANNs) model classified 98 % of samples correctly. The identification of three out of six ions predicted by the ANNs model to be indicative biomarkers that have good predictive performance were identified using MALDI PSD, AP-MALDI MSIMS and LC-ESI-MS/MS. Two of the ions were shown to belong to the same identified peptide, u-l-acid glycoprotein precursor (l, 2) which correctly predicted 95 % (i.e. 45/50) of metastatic melanoma patients.