Evaluation of a GPGPU-based de novo Peptide Sequencing Algorithm

Tandem mass spectrometry (MS/MS) can be used to identify peptides present in a biological sample containing unknown proteins. De novo peptide sequencing aims to determine the amino acid sequence of a portion of a peptide directly from MS/MS spectral data. Unlike spectral cross-correlation methods of peptide sequencing, the de novo approach does not require a complete database of all possible proteins that may be present in the sample. In this work, a de novo peptide sequencing algorithm (denovoGPU) was implemented using general-purpose computing techniques on a graphics processing unit (GPGPU), in order to reduce the runtime of the algorithm sufficiently to complete in real-time during MS/MS data collection. This is a step towards enabling true information-driven MS/MS, where incremental data analysis is used to guide data collection. Given data from an MS/MS spectrum, the algorithm filters the data, generates and scores candidate "sequence tags" (or short amino acid sequences), and ultimately outputs a ranked list of sequence tags. The denovoGPU algorithm was tested on over 380 experimentally obtained MS/MS spectra, whose peptide sequences were validated using the Mascot search engine for mass spectrometry data. The performance of the algorithm was compared to an existing de novo peptide sequencing algorithm (PepNovo) in terms of runtime and sequence tag accuracy. Constraints of the denovoGPU algorithm due to limited GPU memory were identified. By adjusting various parameters of the denovoGPU algorithm, the runtime was reduced to below one second, which is an essential requirement for real-time information-driven MS/MS.