model, as this strategy improves the MS2-based scoring model
for PyProphet and outperforms learning on each single fractions which might lead to low scores on the less populated
fractions at the beginning and end of the gradient. Applying
one scoring model outperforms semi-supervised learning on
each fraction and allows comparison of qualitative and quantitative data across the SEC gradient.
pyprophet score --in= vmatej_I190208_129.osw --level=ms2
The learned weights are saved in the input file, which can
then be used to score all the fractions in parallel.
for run in vmatej_*.osw do pyprophet score --in=${run} -apply_weights= vmatej_I190208_129.osw --level=ms2 done
At this step, all files are merged into a single scoring model.
for run in vmatej_*.osw do run_reduced=${run}r pyprophet
reduce --in=${run} --out=${run_reduced} done
To generate a merged file, the spectral library file is used as
a template. Importantly, it has to be the same library that is
used to perform the quantitation with OpenSWATH.
pyprophet merge –template=HS_decoy_EvosepLib.pqp --out=model_global_2.osw *.oswr
The merged file is used to control false discovery rate at the
peptide and protein level either globally or per file.
pyprophet peptide --context=run-specific --in=model_global_2.
osw
pyprophet protein --context=global --in=model_global_2.osw
The run-specific and global error rates are then transferred
back to each individual file (see Note 24).
for run in vmatej_*.osw do pyprophet backpropagate --in=
${run} --apply_scores=model_global_2.osw done
In the last PyProphet step, the OSW results are exported
for each run and by applying confidence score thresholds. For
SEC data, it is beneficial to export the data with higher FDR
thresholds, as stringent filtering is performed within the
CCprofiler package.
284
Andrea Fossati et al.
for PyProphet and outperforms learning on each single fractions which might lead to low scores on the less populated
fractions at the beginning and end of the gradient. Applying
one scoring model outperforms semi-supervised learning on
each fraction and allows comparison of qualitative and quantitative data across the SEC gradient.
pyprophet score --in= vmatej_I190208_129.osw --level=ms2
The learned weights are saved in the input file, which can
then be used to score all the fractions in parallel.
for run in vmatej_*.osw do pyprophet score --in=${run} -apply_weights= vmatej_I190208_129.osw --level=ms2 done
At this step, all files are merged into a single scoring model.
for run in vmatej_*.osw do run_reduced=${run}r pyprophet
reduce --in=${run} --out=${run_reduced} done
To generate a merged file, the spectral library file is used as
a template. Importantly, it has to be the same library that is
used to perform the quantitation with OpenSWATH.
pyprophet merge –template=HS_decoy_EvosepLib.pqp --out=model_global_2.osw *.oswr
The merged file is used to control false discovery rate at the
peptide and protein level either globally or per file.
pyprophet peptide --context=run-specific --in=model_global_2.
osw
pyprophet protein --context=global --in=model_global_2.osw
The run-specific and global error rates are then transferred
back to each individual file (see Note 24).
for run in vmatej_*.osw do pyprophet backpropagate --in=
${run} --apply_scores=model_global_2.osw done
In the last PyProphet step, the OSW results are exported
for each run and by applying confidence score thresholds. For
SEC data, it is beneficial to export the data with higher FDR
thresholds, as stringent filtering is performed within the
CCprofiler package.
284
Andrea Fossati et al.