1160 Proteomics 2012, 12, 1160–1169 DOI 10.1002/pmic.201100425 RESEARCH ARTICLE Information-dependent LC-MS/MS acquisition with exclusion lists potentially generated on-the-fly: Case study using a whole cell digest of Clostridium thermocellum Peter McQueen 1 , Vic Spicer 2 , Thomas Rydzak 3 , Richard Sparling 3 , David Levin 4 , John A. Wilkins 1,5 and Oleg Krokhin 1,5 1 Manitoba Centre for Proteomics and Systems Biology, Winnipeg, Canada 2 Department of Physics and Astronomy, University of Manitoba, Winnipeg, Canada 3 Department of Microbiology, University of Manitoba, Winnipeg, Canada 4 Department of Biosystems Engineering, University of Manitoba, Winnipeg, Canada 5 Department of Internal Medicine, University of Manitoba, Winnipeg, Canada We have developed a real-time graphic-processor-unit-based search engine capable of high- quality peptide identifications in <500 s per spectrum. The steps of peptide/protein identi- fication, in-silico prediction of all possible tryptic peptides from these proteins, and the pre- diction of their expected retention times and m/z values take less than 5 s per cycle over ∼3000 MS/MS spectra. This lays the foundation for information-dependent acquisition with exclusion lists generated on-the-fly, as the instrument continues to acquire data. While a com- plete evaluation of the dynamic exclusion system requires the participation from instrument vendors, we conducted a series of model experiments using a whole cell tryptic digestion of the bacterium Clostridium thermocellum. We ran a series of five iterative LC-MS/MS runs, adding a new exclusion list at each of four chromatographic “tripping points” – the elution times of the four standard peptides spiked into the sample. Retention times of these standard peptides were also used for real-time “chromatographic calibration.” The dynamic exclusion approach gave a ∼5% increase in confident protein identification (for typical 2 h LC-MS/MS run), and reduced the average number of identified peptides per protein from 4.7 to 2.9. Its application to a two-times shorter gradient gave a ∼17% increase in proteins identified. Further improve- ments are possible for instruments with better mass accuracy, by employing a more accurate retention prediction algorithm and by developing better understanding of the possible chemical modifications and fragmentations produced during electrospray ionization. Keywords: Information-dependent acquisition / Peptide retention prediction / RP HPLC / Shotgun proteomics / Technology Received: August 15, 2011 Revised: December 20, 2011 Accepted: January 1, 2012 Correspondence: Dr. Oleg Krokhin, Manitoba Centre for Pro- teomics and Systems Biology 799 JBRC, 715 McDermot Avenue, Winnipeg R3E 3P4, Canada E-mail: krokhino@cc.umanitoba.ca Fax: +1-204-480-1362 Abbreviations: CPU, central processing unit; CUDA, Compute Unified Device Architecture; GPU, graphic processor unit; HI, hydrophobicity index; IDA, information-dependent acquisition; MGF, Mascot Generic File; SpMV, sparse matrix vector; SSRCalc, Sequence Specific Retention Calculator 1 Introduction Accelerated developments of mass analyzers have rev- olutionized the field of analytical biochemistry. Protein identification, detailed characterization, and quantitation have become a routine procedure for many laboratories around the world; the vast majority of these applications are based on LC-MS/MS analysis of digested protein sam- ples, i.e. a bottom-up approach [1]. Information-dependent acquisition (or data-dependent acquisition (DDA)) is a key C  2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com