Increased Power for the Analysis of Label-free LC-MS/MS Proteomics Data by Combining Spectral Counts and Peptide Peak Attributes

Abstract

Liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based proteomics provides a wealth of information about proteins present in biological samples. In bottom-up LC-MS/MS-based proteomics, proteins are enzymatically digested into peptides prior to query by LC-MS/MS. Thus, the information directly available from the LC-MS/MS data is at the peptide level. If a protein-level analysis is desired, the peptide-level information must be rolled up into protein-level information. We propose a principal component analysis-based statistical method, ProPCA, for efficiently estimating relative protein abundance from bottom-up label-free LC-MS/MS data that incorporates both spectral count information and LC-MS peptide ion peak attributes, such as peak area, volume, or height. ProPCA may be used effectively with a variety of quantification platforms and is easily implemented. We show that ProPCA outperformed existing quantitative methods for peptide-protein roll-up, including spectral counting methods and other methods for combining LC-MS peptide peak attributes. The performance of ProPCA was validated using a data set derived from the LC-MS/MS analysis of a mixture of protein standards (the UPS2 proteomic dynamic range standard introduced by The Association of Biomolecular Resource Facilities Proteomics Standards Research Group in 2006). Finally, we applied ProPCA to a comparative LC-MS/MS analysis of digested total cell lysates prepared for LC-MS/MS analysis by alternative lysis methods and show that ProPCA identified more differentially abundant proteins than competing methods. Liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based proteomics provides a wealth of information about proteins present in biological samples. In bottom-up LC-MS/MS-based proteomics, proteins are enzymatically digested into peptides prior to query by LC-MS/MS. Thus, the information directly available from the LC-MS/MS data is at the peptide level. If a protein-level analysis is desired, the peptide-level information must be rolled up into protein-level information. We propose a principal component analysis-based statistical method, ProPCA, for efficiently estimating relative protein abundance from bottom-up label-free LC-MS/MS data that incorporates both spectral count information and LC-MS peptide ion peak attributes, such as peak area, volume, or height. ProPCA may be used effectively with a variety of quantification platforms and is easily implemented. We show that ProPCA outperformed existing quantitative methods for peptide-protein roll-up, including spectral counting methods and other methods for combining LC-MS peptide peak attributes. The performance of ProPCA was validated using a data set derived from the LC-MS/MS analysis of a mixture of protein standards (the UPS2 proteomic dynamic range standard introduced by The Association of Biomolecular Resource Facilities Proteomics Standards Research Group in 2006). Finally, we applied ProPCA to a comparative LC-MS/MS analysis of digested total cell lysates prepared for LC-MS/MS analysis by alternative lysis methods and show that ProPCA identified more differentially abundant proteins than competing methods. One of the fundamental goals of proteomics methods for the biological sciences is to identify and quantify all proteins present in a sample. LC-MS/MS-based proteomics methodologies offer a promising approach to this problem (1.Aebersold R. Mann M. Mass spectrometry-based proteomics.Nature. 2003; 422: 198-207Crossref PubMed Scopus (5484) Google Scholar, 2.Domon B. Aebersold R. Mass spectrometry and protein analysis.Science. 2006; 312: 212-217Crossref PubMed Scopus (1559) Google Scholar, 3.Cravatt B.F. Simon G.M. Yates 3rd, J.R. The biological impact of mass-spectrometry-based proteomics.Nature. 2007; 450: 991-1000Crossref PubMed Scopus (554) Google Scholar). These methodologies allow for the acquisition of a vast amount of information about the proteins present in a sample. However, extracting reliable protein abundance information from LC-MS/MS data remains challenging. In this work, we with the analysis of data using bottom-up label-free LC-MS/MS-based proteomics to the that proteins are enzymatically digested into peptides prior to query by the LC-MS/MS up protein by mass Scopus Google and that are the of One in the bottom-up approach to proteomics is that information directly available from the LC-MS/MS data is at the peptide level. a protein-level analysis is desired, as is the with LC-MS peptide-level information must be rolled up into protein-level information. counting Yates 3rd, J.R. for and of relative protein abundance in PubMed Scopus Google Scholar, Mann M. protein abundance for of protein amount in proteomics by the of peptides PubMed Scopus Google Scholar, R. protein the relative of and 2007; PubMed Scopus Google Scholar, proteomic and of the 2007; PubMed Scopus Google Scholar, analysis of spectral count data in label-free PubMed Scopus Google Scholar, B. of relative abundance derived from peptide ion and counting for quantitative proteomic analysis using PubMed Scopus Google is a and used of peptide-protein for LC-MS/MS in and may to the of to peptide in a or using peptide platforms Yates 3rd, J.R. approach to mass data of peptides with in a protein Mass PubMed Scopus Google and of proteins by PubMed Scopus Google for the identified peptide in to In the of to peptides to a the spectral count used mass and component peak peptide-protein peptide-protein is with the abundance of a protein Yates 3rd, J.R. for and of relative protein abundance in PubMed Scopus Google Scholar). In spectral counting or are used as a for protein counting methods in protein abundance and proteins in However, methods of information available from in the LC-MS and this to of spectral count mass and principal component analysis peptide peak alternative peptide-protein peptide-protein in the LC-MS to peptide ion are to in protein abundance M. of proteins and by mass spectrometry or 2003; PubMed Scopus Google Scholar, for proteomic and using chromatography-tandem mass PubMed Scopus Google Scholar). LC-MS that to peptides and quantitative of as area, or a promising alternative to spectral counting methods. These methods in using Mann M. peptide mass and protein PubMed Scopus Google Scholar). However, in the label-free analysis of proteomics in peak and are a In and quantification of LC-MS peptide peak may be using peak platforms G.M. a for proteomic 2006; PubMed Scopus Google Scholar, M. M. M. M. R. M. of for the analysis of protein using 2006; PubMed Scopus Google Scholar, M. M. for mass and of mass spectrometry 2007; PubMed Scopus Google Scholar). of protein abundance may be by up into protein-level information G.M. a for proteomic 2006; PubMed Scopus Google Scholar, a statistical for quantitative analysis of PubMed Scopus Google Scholar, quantification of mass spectrometry data for proteomic PubMed Scopus Google Scholar). peptide-protein the of all peptides to a protein to a protein-level of the of for protein the performance of methods may the are by in and peptide peak attributes. These are as the the of methods. existing peak and quantification platforms to in data with G.M. a for proteomic 2006; PubMed Scopus Google Scholar, a statistical for quantitative analysis of PubMed Scopus Google Scholar, M. M. methods for analysis of PubMed Scopus Google with dynamic and ion are that must be in to in protein-level abundance peptide-protein that for this and the in a However, in the of is in the existing peptide-protein for In this work, we propose ProPCA, a peptide-protein for efficiently extracting protein abundance information from bottom-up label-free LC-MS/MS ProPCA is easily that is to component analysis of Google Scholar). ProPCA and data to of relative protein ProPCA in data in a of both and methods. In ProPCA to the of the available If the data are in the are ProPCA is to spectral the other is in the data the ProPCA is a of and spectral the are to the of spectral and peptide to protein we the performance of ProPCA using a data set from the LC-MS/MS analysis of protein standards proteomic dynamic range standard set in by M. and The Association of Biomolecular Resource Facilities Proteomics Standards Research by and show that ProPCA outperformed other existing methods by The of ProPCA is by the quantification used to and we show that ProPCA to used with alternative quantification Finally, we applied ProPCA to a comparative LC-MS/MS analysis of digested total cell lysates prepared for LC-MS/MS analysis by alternative lysis methods. We show that ProPCA identified more differentially abundant proteins than competing methods. was with and a was used to a of and a and a both with protein to of the protein was the with the the The was up and using the and and the with a of in The LC-MS/MS in and in The the for in in for and in The was introduced into mass with a by The was set to and the of the was set to was in the mass at mass and ion was by for the for of the and mass The mass was set for and at was for for of the that for to from The UPS2 as using a LC-MS of the UPS2 was by LC-MS/MS to was both the UPS2 standards and the cell the data by the mass and to the and for using the the UPS2 the was a in the UPS2 protein and proteins from of for all proteins in the the the was a in total and proteins from the from the the from and and and with set as or peptide mass to and a ion mass to used in all mass peptide and up to used in all count information was from in We the of a protein in a by counting the of in the to peptides that to the protein may that a peptide to more than the UPS2 standard a was of identified peptides to in the data of identified peptide to may to in In a peptide was to we the peptide to a protein from the of may into the was the of peptide-protein this more of peptides to proteins is this was the of this The data protein including from for the UPS2 and information for the UPS2 data is in a to peptides with than spectral in we proteins that identified by at The was as the of peptide from a by the total of protein and this was to a to of with mass spectrometry for protein the 2003; PubMed Scopus Google and of peptide and protein from mass spectrometry and the PubMed Scopus Google the was for both the UPS2 and We used M. M. M. M. R. M. of for the analysis of protein using 2006; PubMed Scopus Google Scholar, M. M. for mass and of mass spectrometry 2007; PubMed Scopus Google Scholar, M. M. in mass and analysis and in proteomics PubMed Scopus Google and LC-MS to from the platforms peak and are of for a peptide in the peptide was identified at the by information from other The peak M. M. M. M. R. M. of for the analysis of protein using 2006; PubMed Scopus Google Scholar, M. M. for mass and of mass spectrometry 2007; PubMed Scopus Google Scholar, M. M. in mass and analysis and in proteomics PubMed Scopus Google the LC-MS a using we the LC-MS/MS data into R. M. B. B. R. R. R. Aebersold R. of mass spectrometry data and to proteomics PubMed Scopus Google using the available at we In the we and in the the UPS2 we set to and to with for all other the used for the we set to the LC-MS peak from and the the for the UPS2 we set to to to to and to used for all other We used the for the cell we set to and to Finally, to in the LC-MS to peptides identified with the of the both the UPS2 and the used for the of and of that with M. M. in mass and analysis and in proteomics PubMed Scopus Google than The data the information for ProPCA and the other we The data information from and for all UPS2 and samples. was to information using the LC-MS We and and the LC-MS using for of the and the of analysis using the The LC-MS of using peptide and we with and The for peptide was and the the for analysis of the The data the including for the UPS2 (the with the LC-MS the of we to to the of and the of and the ProPCA of relative protein the for the peptide to the protein in the sample. We that are in that a total of proteins and that peptides to the Thus, for the and and We the abundance of the protein in the sample. and protein abundance a statistical the or and protein is by is the of and and are that in the is up to is to the that ProPCA of relative to protein abundance and is In the of a peptide-protein is to for and M. The ProPCA and the and are as of the In other the ProPCA are the that in and with to data are a of the are available for all and are the ProPCA to the component by the data for protein In the of ProPCA are by the is such that is this problem may be by using a using Scholar, Scopus Google Scholar). and the ProPCA are to M. R. methods for PubMed Scopus Google Scholar). we a of for ProPCA a protein is be the of to the and is to a to in the data to a in the the as and that for is to of with In for be in a and is to the component to the data both the and the The for and ProPCA is as The data for the of the is data for the be the of the data for the by is and the is the ProPCA for the is easily and in is in with in to The in with and in of lysis as as and The lysis used for The and in The in at for and for was at for at to cell The was to with and was to The lysates and for at to and to was to and the lysates and for at in the was to a of and for in the at The lysates with of the lysates to and to of was to to of was to to to and to a total of in The for at and to The by mass spectrometry as The for ProPCA, and is in the data and is available at The data set used to the performance of ProPCA was derived from the LC-MS/MS analysis of of the UPS2 proteomic dynamic range standard by The UPS2 standard set proteins with a dynamic range of of to the The used in analysis of of the UPS2 by a and of data from LC-MS/MS available for the UPS2 and the of the UPS2 standard a protein dynamic range of more than of ProPCA and that must be from the LC-MS/MS platforms that this are We used and Aebersold R. statistical to the of peptide by and PubMed Scopus Google to identify peptides and proteins by and and to In analysis of the UPS2 peptides to of the proteins in the UPS2 standard set identified or used to a dynamic range this was the of this The identified proteins abundance proteins in the UPS2 standard set with from to for we used M. M. M. M. R. M. of for the analysis of protein using 2006; PubMed Scopus Google Scholar, M. M. for mass and of mass spectrometry 2007; PubMed Scopus Google in LC-MS peptide peak LC-MS the from and to identified We to the with LC-MS/MS data and and as the data the data for ProPCA may be as a a is in spectral count and information about the protein for LC-MS/MS from the UPS2 standard data In a data with for LC-MS/MS in the UPS2 data is available for and this was used in statistical is as a for all proteins from all for both the UPS2 and are available in the data in the of and may be easily with the statistical in that the was to a to the peptide in the sample. may be the peptide is present in the or of in the the other in a of and are available for the is are in the peak is to and quantify information from other samples. the data in the of peptide-protein is to and into a for that protein and information for from LC-MS/MS protein to peptides to in the UPS2 LC-MS/MS data the of of the of the are by protein The to peptides to in the UPS2 LC-MS/MS data the of of the of the are by in a may be to the LC-MS/MS data to the and in the that the of is with the of protein abundance R. protein the relative of and 2007; PubMed Scopus Google Scholar, proteomic and of the 2007; PubMed Scopus Google and this is with the of we to to the of the of and protein all proteins identified in the UPS2 was We that other that are with the of protein abundance Mann M. protein abundance for of protein amount in proteomics by the of peptides PubMed Scopus Google or protein abundance Yates 3rd, J.R. for and of relative protein abundance in PubMed Scopus Google In we the However, the was this may be to the dynamic range in the UPS2 standard of we that the of are with the of protein abundance peptides identified in UPS2 standards and that the of are with the of protein abundance and for a proteins and the data for all identified peptides and proteins in the UPS2 about in and we a to both and the we with and In to to the may be to and peptide peak and samples. analysis of spectral count data in label-free PubMed Scopus Google Scholar, J.R. of label-free methods for proteins by PubMed Scopus Google Scholar, B. and protein in label-free 2006; PubMed Scopus Google Scholar, B. analysis of in 2006; PubMed Scopus Google is the of is the abundance of a protein relative to total protein abundance or to the abundance of proteins in a biological sample. In UPS2 standard we as of the of the protein We in the cell data at the to cell from of in the cell protein abundance is a of may be to the of and peptide peak is a of combining of protein However, in with ProPCA and the other methods for peptide-protein we that to in protein approach for to and the of peptide that and standard in this we that the ProPCA and protein abundance in the UPS2 standard set with the we a in with protein abundance alternative peptide-protein used We that the in data may be to the in data and in and standard we in analysis However, into for ProPCA and other peptide-protein may be counting and methods for protein quantification are by the that are with protein abundance in the UPS2 and both with protein ProPCA are derived by the that and with the of protein ProPCA is for the of relative protein In the data the ProPCA for a protein are to the principal component of Google of the protein data are ProPCA are using a using Scholar). ProPCA provides of the relative protein abundance of identified protein in sample. with data protein are to ProPCA are up to In the of this may be in to protein However, of the of statistical to this is for a protein is differentially abundant samples. In to of relative protein the ProPCA to the spectral count and peptide (the of In the data ProPCA is to principal component the relative of spectral and the peptides to the ProPCA that spectral or the peptide a in the ProPCA the of is the may offer into the spectral and peptide in the ProPCA The for the UPS2 standard data are in and we the performance of ProPCA with that of and existing peptide-protein that was by G.M. a for proteomic 2006; PubMed Scopus Google Scholar). to as alternative protein protein-level are by by the for the peptide to peptide a peptide was are to peptide to for are by the of all peptide The protein abundance and ProPCA, and was for identified protein in the UPS2 The ProPCA and protein abundance was and the with protein abundance was and that ProPCA with protein abundance than and of the protein abundance for proteins are in the data for all identified proteins in the UPS2 with the of protein abundance the of ProPCA protein or relative that are proteins as as remains and that we in proteomic and of the 2007; PubMed Scopus Google Scholar). we the of ProPCA to the abundance of a protein We the of ProPCA, and to with protein used in with ProPCA, and we UPS2 with protein These for protein identified in the UPS2 standard data set and of protein to a the of is of the of to with protein In a more of the performance of may be by the of in this as to the of that are or other in this a The is in more in the a of the the of the UPS2 and of the identified we ProPCA, and of protein abundance all with protein amount or we for and identified with protein with with We a with than to be protein abundance a was for all of protein in the UPS2 In for of the We the of for methods and that ProPCA in of and of for and that ProPCA used with is at differentially abundant proteins in the UPS2 with and In the of may be the data a B. of Scholar). and in by with more However, the ProPCA, and data as by the for analysis of for Google for ProPCA, and are all the other in analysis of the performance of ProPCA, we in the we in the of from the and the relative performance of ProPCA as with and the data may be this the from the from with the protein abundance are the that the is the the are of about the the in we used a to the of and we that this is for identified protein in the UPS2 we of the and the protein abundance and to of We for identified protein and method, the We this for Thus, in for are to we that is the that the for is to the that analysis in is We the of with LC-MS/MS of the data and a of is in LC-MS/MS the and in alternative to may be more M. and quantitative proteomics PubMed Scopus Google Scholar). In ProPCA may be used in with for the statistical analysis of relative protein abundance that a However, that to with the UPS2 standard we that the is a for the performance of ProPCA of ProPCA are more with protein abundance than as in the we that using offer a reliable of the comparative performance of ProPCA, and used with more methods. be that ProPCA be used in with the J.R. of label-free methods for proteins by PubMed Scopus Google or other that analysis of spectral count data in label-free PubMed Scopus Google for a is available at in a for may be used in of the and to a of the methods by analysis of spectral count data in label-free PubMed Scopus Google we a and Google to the of alternative approach to estimating is We the approach of the with LC-MS/MS proteomics approach more the available In the UPS2 the was in the that peptides identified by to in the LC-MS at a In data set of the UPS2 the was in the cell data the was (the was by the total of to peptide ion LC-MS by the of the total of and the total of peptides identified by this was by to a The in the data was and to the of cell the performance of ProPCA in a we from the UPS2 standard data set to (the We data with and of data and method, we of protein abundance as in the to the of ProPCA at We the with data set to the that ProPCA and the range of a in a to the that of the cell may are in for data with may the the statistical for protein in bottom-up PubMed Scopus Google we that may offer into the performance of and of in the analysis of the performance of the other with the of to of the the of LC-MS/MS be of the of ProPCA to the data we used LC-MS to alternative from the UPS2 standard data the data and the data in we and ProPCA The of ProPCA, and with protein abundance was and We and to the are in the using LC-MS to the of analysis using ProPCA outperformed and However, the performance was as The of ProPCA and using the LC-MS with using to the that from the LC-MS as with protein abundance as from for the LC-MS the ProPCA and protein abundance was and the with the of protein abundance was and that the for are the as in the analysis was used to is we used the to in both the other the for ProPCA and both than in the analysis using the is by the that the and protein abundance is for the that in the data the and protein abundance was In the the and protein abundance was the is all We the and protein abundance to the was more for the The the and protein abundance was is from the and protein the performance of ProPCA in with other we of ProPCA to the of LC-MS/MS analysis of total cell of using and prepared for LC-MS/MS In the lysis in the other was than this the a with was applied to the of and and to the of protein by the for Google Scholar). analysis by and was used to In data from LC-MS/MS available from In the cell peptides and proteins in total identified by all spectral a of spectral and peptide and protein information for the cell ProPCA or other protein abundance we that into protein may be by total spectral peptide and protein information for the methods. In the data to lysis was the of spectral peptides and proteins identified was and In the data to lysis was the of spectral peptides and proteins identified was and These protein is by LC-MS/MS analysis of prepared with the of the lysis was standard to spectral peptides and proteins identified and of proteins identified in at in identified in all for of all proteins identified in the identified in all and of all proteins identified in the lysis identified in all The for peptide as of all peptides identified in the lysis identified in all and of all peptides identified in the lysis identified in all We that be in of spectral and peptide and protein information for cell spectral in a We ProPCA, and protein abundance for identified protein and used to identify proteins that differentially from the using more to for in protein abundance may be of in the present we used as to a more to the of ProPCA, as with and we that the and standard of ProPCA was using all identified proteins in the data is and standard is for to M. and quantitative proteomics PubMed Scopus Google Scholar, B. for for PubMed Scopus Google Scholar). ProPCA, of the proteins to be at the level. and and proteins to be at the Thus, as with the UPS2 ProPCA more proteins with than the These for to as are a of are for we the the a and approach to R. B. to identify proteins and a of to for such as the B. of the of (the than the of this to be The the other is a used and easily statistical for the at a level. the and proteins to be using ProPCA, and ProPCA more proteins at the we that a of than ProPCA (the ProPCA is and are than may be to the and in the cell data and that the in the data was this we that ProPCA more to differentially proteins than We that with ProPCA show performance and is by analysis of the UPS2 the in that the lysis may to the of more proteins by LC-MS/MS analysis than the lysis the ProPCA that the lysis or to the of of the in protein by lysis we analysis by of the using the proteins identified by ProPCA that at a to the and the are in that the of into the lysis to of proteins of and and proteins of other in to of We this of as as and proteins to the of to and with and such as and proteins and cell as as proteins using the lysis the of the of for this we that the of in of protein and the ProPCA analysis the of of cell or lysis to proteins of more is in the analysis of bottom-up LC-MS/MS proteomics We ProPCA, a for peptide-protein roll-up, and that ProPCA are more with the of protein abundance than derived using other peptide-protein we that ProPCA to in protein abundance than competing In be to more than and we ProPCA to in this In to the ProPCA in the analysis of the UPS2 we that ProPCA identified more proteins than other applied to the cell with using of in quantitative of a of the up the and platforms in mass spectrometry analysis allow for more of a and more of However, in the for of the protein by or lysis be to that with ProPCA a of the The to the existing of ProPCA and are derived from label-free proteomics However, in ProPCA may be used for peptide-protein in the analysis of proteomics that methods. We for with with