Tutrial 7: Autmated SRM data analysis using mprphet mprphet is a statistical tl which can be emplyed t achieve autmated high-cnfidence identificatin f peptides. It has three functinalities: First, mmap cmbines the raw data (mzxml files) with the meta data prvided in the transitin input file. Secnd, mquest detects all peak grups amng the recrded transitins and scres all detected peak grups fr a number f characteristics (e.g. shape crrelatins, etc.). Third, mprphet cmbines the individual peak grup scres int a single discriminant scre (d_scre) fr ptimal recvery f true signals, while cntrlling the FDR. Decy transitin grups are used as negative cntrls. They represent peptide species that are absent frm the bilgical samples, and help t parameterise the null distributin fr estimating the sensitivity and the FDR. Finally, all detected peak grups fr each recrded transitin grup are ranked based n their discriminant scre and reprted. Nte that mprphet relies n prbabilistic mdelling. It subdivides each dataset int a training set and a test set, in rder t first train a classifier and then apply it t the whle dataset. Since a training set is chsen at randm, results frm repeated analyses f the same data may vary slightly. mprphet can use tw different types f decy transitins: Empirical decy transitin grups are included in the measurements. They can be created e.g. by adding a randm integer t the precursr and fragment in masses, respectively. The empirical decy transitin apprach is highly recmmended fr wrkflws that lack reference peptides. The disadvantage f this apprach is that the empirical decy transitin grups have t be added t the transitin list befre the SRM measurements, which increases the amunt f transitins t be measured. Synthetic decy transitin grups are derived pst-acquisitin, by shifting the retentin time f the endgenus transitin peak signals by a factr defined in the parameter file (0.5 is the recmmended value). The synthetic decy apprach is recmmended fr wrkflws that include istpe-labelled internal standards in the samples. The advantage f this apprach is that the decy transitin grups d nt have t be measured. mprphet ffers several pre-defined wrkflws, which reflect the design f the experiment and define specifics f the decy transitin apprach. In this tutrial we will apply the tw mst cmmn appraches t ur case study. Fr an verview f all wrkflws see the table at the end f this tutrial. Wrkflw SPIKE_IN + synthetic decys Parameter file: param_aqua_heavy_stringent_ref_synthetic.def Target transitins measured fr light and heavy peptides Reference peptides are heavy peptides and we assume they are always present Synthetic decy transitin grups are generated frm the data Wrkflw LABEL_FREE + empirical decys Parameter file: param_light_nref.def Target transitins measured fr light peptides N reference peptides Empirical decy transitin grups measured alng with target transitin grups 1
Open the mprphet virtual machine Duble click n the.vmx file t pen the mprphet virtual machine (VM) and lgin if necessary (username: mprphet, passwrd: mprphet). Enable flder sharing in the virtual machine settings Fr the VMware player (Windws) the flder sharing is dne in either f the fllwing ways, depending n the versin: Player à Manage à Virtual machine settings à Optins à Shared flders. Select Always enabled and Add the flder Tutrial- 7_mPrphet. VM à Settings à Optins à Shared flders. Select Always enabled and Add the flder Tutrial-7_mPrphet. Fr VMware Fusin (Mac) the shared flder is added under: Virtual Machine à Settings à Sharing. Turn "ON" the "Shared flders" ptin, click n the "+" buttn, brwse t the flder Tutrial-7_mPrphet and click "Add". On the VM yur shared flder can be accessed thrugh: Places à Cmputer à File Systems à mnt à hgfs à Tutrial-7_mPrphet T adjust the keybard t yur preferred language settings: System à Preferences à Keybard Flder system t brwse files Terminal / Cmmand line Ggle Chrme brwser Change keybard language Exprt transitin infrmatin frm Skyline using a custm reprt frmat On yur laptp (nt VM), pen the file SRMcurse_20130717_iRT.sky frm the flder Tutrial-5_Scheduling. Exprt transitins and ther relevant infrmatin in a custm reprt frmat: File à Exprt à Reprt à Imprt à Select mprphet.skyr à mquest_mrmprphet nw appears in the list. Inspect it by clicking the Preview buttn. One clumn has t be added t this new reprt: Edit list à Select mquest_mrmprphet reprt à Edit à Rename it t mquest_mrmprphet_irt and add the clumn RetentinTimeCalculatrScre à click 2x OK. Save the resulting transitin list as mprphet _20130717_label.csv int the flder mprphet_20130717_label in Tutrial-7_mPrphet. 2
Refrmat Skyline reprt using mgen On the VM, pen the terminal and change t the fllwing directry: cd /mnt/hgfs/tutrial-7_mprphet/mprphet_20130717_label Refrmat line breaks f the transitin list t UNIX frmat: Install the tfrds: sud apt-get install tfrds (pwd is mprphet ). This cmmand requires a wrking internet cnnectin. Run: frmds mprphet_20130717_label.csv Run mgen: mgen.pl -SKY mprphet_20130717_label.csv -num_decys 0 We set 0 decys, because we d nt want t generate decys here. The utput file ends with _s01_decy. On yur laptp (nt VM), pen the newly generated file in Excel and rerganise it such that nly the fllwing clumns remain (headers f the required clumns are given in brackets): 1. Precursr in mass (Q1) 2. Fragment in mass (Q3) 3. Prtein name (prtein_name) 4. Unmdified peptide sequence (stripped_sequence) 5. Precursr charge state (prec_z) 6. Istype f the peptide, e.g. heavy r light (istype) 7. Fragment in type, e.g. b- r y-in (frg_type) 8. Fragment number (frg_nr) 9. Fragment in charge state (frg_z) 10. Relative library intensity f transitins within ne transitin grup, nrmalised t 100. (relative_intensity) 11. Expected retentin time f the peptide: Rename PredictedRetentinTime t Tr_recalibrated. (Tr_recalibrated) 12. irt: Rename RetentinTimeCalculatrScre t irt (irt) 13. Decys: In the label-wrkflw n decys are needed, hence this clumn can be deleted. Otherwise enter 1 fr decys and 0 fr nn-decys. (decy) 14. Delete all ther clumns. Save as tab-delimited file: mprphet_20130717_label_md.txt. - On mac save as MS-DOS txt. Rename ending t.xls: mprphet_20130717_label_md.xls. Refrmat line breaks t UNIX: frmds mprphet_20130717_label_md.xls Delete the ther.csv, xls and.txt files frm the flder. Run mprphet t identify target prteins T run mprphet, make sure yu have all data files (.mzxml), the transitin list (.xls), and the parameter file (.def) in flder mprphet_20130717_label. On the VM, run all three functinalities f mprphet (mmap, mquest and mprphet) tgether using the minteract.pl cmmand: minteract.pl -mmap -mquest -mprphet -wrkflw SPIKE_IN - mmap_machine QTRAP -mmap_cycle_time 2 -def param_aqua_heavy_stringent_ref_synthetic.def Ntes This cmmand takes a while and might give a lt f errr messages. Please wait until it is finished befre yu click anything. Yu can mnitr the prgress by lking at the newly generated files in the flder mprphet_20130717_label. Fr TSQ data, the parameters -mmap_machine and - mmap_cycle_time d nt have t be defined. If yu wuld like t repeat the mprphet analysis n the same data, use the -frce ptin at the end f the minteract cmmand line t enfrce the re-analysis f files that have already been prcessed. 3
Tip! It is better t type these cmmands manually int the terminal rather than t cpy and paste them, t avid incmpatibilities between DOS and UNIX. Tip! T see a list f all pssible parameters fr mprphet, run: minteract.pl -manual Tip! minteract crdinates mmap, mquest and mprphet and allws a ne cmmand analysis. Hwever, the three cmpnents can als be run separately (see mprphet manual n the www.mprphet.rg website fr mre infrmatin). Inspect mprphet results mprphet results are written t the flder mprphet_20130717_label where yu ran the cmmand. The fllwing result files are imprtant: mprphet.pdf This file cntains histgrams f all the different sub-scres and the verall discriminant scre stratified by target and decy peak grups (left figure) as well as the estimated sensitivity and errr rate fr discriminant scre cut-ffs (right figure). à Inspect the separatin between decy and target peak grups and the sensitivity (s-value) and FDR (q-value). The s-value represents the expected prprtin f true psitive peak grups dependent n the discriminant scre cut-ff and the q-value the expected prprtin f false psitives dependent n the discriminant scre cutff. The aim is t define a discriminant scre cut-ff resulting in a high sensitivity at a lw FDR. Nte that the s-value des nt reach 1 because the retentin time peptides are excluded fr the statistical mdels, but later n cunted twards the ttal. mprphet_raw_stat.xls This file prvides the discriminant scre cut-ff fr every pssible FDR. It is recmmended t select a d-scre cut-ff that crrespnds t an FDR f 1%. à In ur case, accrding t mprphet, the FDR f the cmplete dataset withut applying any cut-ff is already very lw (0.2%) and n cut-ff can be applied based n the FDR. mprphet_all_peakgrups.xls This file cntains all the peak-specific scres extracted and calculated frm the SRM traces acrss all the samples. We will need nly a few f the clumns t cntinue (see belw). shtml files These files can be pened frm within a web brwser and allw the visualisatin f all the extracted peak grups fr each SRM run (see next sectin). 4
Visualise the mprphet results On the VM, cpy yur data flders t the VM hme directry: cp -R /mnt/hgfs/tutrial-7_mprphet/mprphet_20130717_label/ //hme/mprphet/mprphet_data_analysis/ There is a space befre the //. D nt change the name f the flders anymre. On the VM, pen Ggle Chrme (link n Desktp) and navigate t yur mprphet results using the fllwing URL: http://lcalhst/mquest-web/mprphet_data_analysis/ Here yu can find fr each f the 9 input data files (mzxml) a crrespnding shtml file, which yu can nw pen and brwse. Click thrugh the peptides f a few samples and inspect the peak picking by lking at the plts and the scres in the table. (Decy peptides are marked by a 1 in the right clumn.) Here are a few pints that yu shuld pay attentin t: Which d-scre cut-ff is apprpriate fr the current data set? At which d-scre d the first decys appear? Fr which precursr was the wrng peak grup picked and why? Cmpare the scres and plts f the secnd and third peak grup by clicking n the peak grup rank in the table. Are there interfered transitins? Prcess mprphet results The mprphet scres fr each precursr can be fund in the file mprphet_all_peakgrups.xls. Befre cntinuing with dwnstream analysis, refine the table in Excel: Remve irt peptides and decys (decy clumn à TRUE). Remve all peak_grup_rank >1 Remve peaks with d-scre lwer than the d-scre cut-ff yu selected frm manual inspectin in the visualisatin step abve. Save as mprphet_all_peakgrups_label.xlsx. 5
mprphet results fr identificatin The m-scre represents the FDR f each identified peak grup. mprphet results fr quantificatin mprphet utputs als cntain all necessary infrmatin fr quantificatin. Yu can either directly extract quantitative infrmatin fr each precursr (e.g. light_heavy_rati_ttalxic) r extract quantitative infrmatin fr each transitin individually (e.g. abs_area_cde_target and abs_area_cde_ref). Dwnstream statistical analysis can be dne e.g. using the sftware SRMstats, which will be discussed in detail tmrrw. Label-free mprphet analysis Repeat the mprphet analysis fr the label-free dataset in the sub-flder mprphet_20130717_label-free. Start with the Skyline file SRMcurse_20130717_label-free_decys.sky frm the flder Tutrial-7_mPrphet which cntains a decy transitin grup fr every target transitin grup. Nte! These decys were added in Skyline thrugh Edit à Refine à Add decy peptides à 30 decy precursrs (Skyline autmatically excludes irt peptides fr decy generatin) à Decy generatin methd: Randm mass shift. We are nt generating these decys urselves in this tutrial, because randm m/z shifts are btained every time decys are generated and thus will nt fit t the acquired data anymre. Next t the m/z f each Precursr and transitin the mass shift f the decy is indicated in brackets. Exprt a reprt using the mquest_mrmprphet_irt reprt frmat: mprphet_20130717_label-free.csv. On the VM, in the terminal change the wrking directry: cd /mnt/hgfs/tutrial-7_mprphet/mprphet_20130717_label-free Cnvert line breaks t UNIX frmat: frmds mprphet_20130717_label-free.csv Run mgen t refrmat it: mgen.pl -SKY mprphet_20130717_label-free.csv -num_decys 0 Mdify the utput as described fr the label-based analysis, but remve the irt clumn and add a decy clumn which cntains 0 fr target peptides and 1 fr decys (indicated in prtein_name clumn). Save as (MS-DOS) tab-delimited file: mprphet_20130717_labelfree_md.txt. Rename ending t.xls: mprphet_20130717_label-free_md.xls. Refrmat t UNIX: frmds mprphet_20130717_label-free_md.xls Delete the.csv and the.txt files frm the flder. Run mprphet: 6
minteract.pl -mmap -mquest -mprphet -wrkflw LABEL_FREE - mmap_machine QTRAP -mmap_cycle_time 2 -def param_light_nref.def Cpy the results t the virtual machine fr visualisatin: cp -R /mnt/hgfs/tutrial-7_mprphet/mprphet_20130717_label-free/ //hme/mprphet/mprphet_data_analysis/ Inspect the peak picking in the shtml-files. Refine the mprphet_all_peakgrups.xls as described abve and save as mprphet_all_peakgrups_label-free.xls. Exercises 1. Which are the scres that mprphet takes int accunt fr the discriminatin f true and false peak grups fr the label-based and the label-free wrkflw, respectively? 2. What is the typical range in delta_irt f the best-scring peak grup? 3. Which d_scre cut-ff wuld yu chse t get gd results fr the label-based and label-free analysis, respectively? Which FDR and sensitivity wuld yu get accrding t the suggested cut-ff (mprphet_raw_stat.xls)? Is the FDR estimated by mprphet what yu wuld expect? Why/why nt? 4. Lk at a few examples where mprphet picked the wrng peak and try t explain why this happened. 5. Hw wuld the results change if yu used as an input fr mprphet a transitin list which has been refined befre in Skyline (i.e. cntaining nly nn-interfered transitins)? Acknwledgements and References The descriptins f this tutrial were adapted frm a manuscript which will sn be published (Surinva et al., Nature Methds 2013, in press). Many thanks t Silvia Surinva and Ruth Hüttenhain fr giving us access t the manuscript! Fr mre infrmatin check the mprphet website (www.mprphet.rg) and the fllwing publicatin: Reiter, L., Rinner, O., Pictti, P., Hüttenhain, R., Beck, M., Brusniak, M.-Y., Hengartner, M.O., and Aebersld, R. (2011). mprphet: autmated data prcessing and statistical validatin fr large-scale SRM experiments. Nature Methds 8, 430 435. We wuld like t thank Prime-XS and SystemsX fr supprting the. 7
mprphet wrkflws and assciated parameter files The fllwing table describes different experimental designs fr SRM experiments and suggests the wrkflw and parameter file required fr the mprphet analysis f SRM data derived frm the different experimental designs (Surinva et al., Nature Prtcls 2013, in press). The parameter file fr all the wrkflws can be fund in the directry /usr/lcal/apps/bignsys/mquest/cnf/ accessible via the mprphet virtual machine. Experimental design Wrkflw Parameter file Target transitins measured fr light Reference peptides are heavy. Empirical decy transitin grups measured fr light peptides. Target transitins measured fr light Reference peptides are heavy. Synthetic decy transitin grups are generated frm the data. Target transitins measured fr light Reference peptides are light. Empirical decy transitin grups measured fr heavy peptides. Target transitins measured fr light Reference peptides are light. Synthetic decy transitins generated frm the data. Target and empirical decy transitins measured nly fr light peptides. Target and empirical decy transitins measured nly fr light peptides. Target and empirical decy transitins measured fr light and heavy peptides. Reference peptides derived frm metablically heavy labeled sample (e.g. SILAC/N15). Target transitins measured fr light Reference peptides derived frm metablically heavy labeled sample (e.g. SILAC/N15). Synthetic decy transitins generated frm the data. Target and empirical decy transitins measured fr light and heavy peptides. Reference peptides derived frm a light (nn-labeled) sample. Target transitins measured fr light Reference peptides derived frm a light (nn-labeled) sample. Synthetic decy transitins generated frm the data. SPIKE_IN SPIKE_IN INVERTED_SPIKE_IN INVERTED_SPIKE_IN LABEL_FREE LABEL_FREE LABEL LABEL LABEL LABEL param_aqua_heavy_stringent_ref.def param_aqua_heavy_stringent_ref_ synthetic.def param_aqua_light_stringent_ref.def param_aqua_light_stringent_ref_ synthetic.def param_light_nref.def param_heavy_nref.def param_silac_heavy_ref.def param_silac_heavy_ref_synthetic.def param_silac_light_ref.def param_silac_light_ref_synthetic.def 8