ஜர்னல் ஆஃப் புரோட்டியோமிக்ஸ் & பயோ இன்ஃபர்மேடிக்ஸ்
திறந்த அணுகல்
ஐ.எஸ்.எஸ்.என்: 0974-276X
ஐ.எஸ்.எஸ்.என்: 0974-276X
Vishna D Nadarajah, Bart JA Mertens, Hans Dalebout, Marco R Bladergroen, Sharmini Alagaratnam, Penny Garrood, Kate Bushby, Volker Straub, André M Deelder, Johan T den Dunnen, Gert-Jan B van Ommen, Peter AC ‘t Hoen and Yuri EM van der Burgt
The accuracy of Mass Spectrometry (MS)-based analysis of peptides in complex biological mixtures improves upon using high resolution instrumentation. However, high resolution content poses challenges to data processing and statistical analysis. Here, three different data handling strategies were evaluated with respect to classification performance using a well-defined cohort of serum samples from Duchenne Muscular Dystrophy (DMD) patients and controls. For this purpose, serum samples were purified using a solid-phase extraction (SPE) protocol based on Reversed-Phase (RP) C18 magnetic beads. Isotopically-resolved peptide profiles were acquired on a Matrix-Assisted Laser Desorption/Ionization Time-of-Flight (MALDI-TOF) mass spectrometer and examined by either using the full mass spectrum or after selecting peaks between 1000<m/z<4000 followed by data filtering or data integration. To identify discriminative peptides, linear Logistic Regression Analysis (LRA) with double-cross validation was applied for each method. The data integration strategy resulted in the lowest classification error rate while use of the filtered or full profile data gave higher error rates. From this it was concluded that peak selection methods may increase the discriminative power, however with the potential downside of loss of potentially interesting peptides. Seven peptides were found by all three methods when considering the top 15 discriminating peptides. Correlation analysis of discriminative peptides showed strong associations between peptides of different m/z-values, suggesting that the list of discriminative peptides reflected a smaller group of proteins. Validation studies using larger patient cohorts are required for further statistical evaluation of these results.