Quickstart

Here, we briefly describe post-processing a PIN file, test.pin, using a deep neural network (DNN) as the classifier during semi-supervised learning.

Recalibrating PSMs

After installation, the main ProteoTorch module, analyze, may be run from the command line using:

proteoTorch --pin test.pin --output_dir testOutput --method 3 --numThreads 10

In the above, --method specifies a DNN classifier, --output_dir specifies the directory to write results to, and --numThreads specifies the number of CPU threads to use for all parallelizable ProteoTorch computation (discussed further in the next section).

When finished, the recalibrated PSM scores will be written to the tab-delimited file testOutput/output.txt. The first few lines of an example output file are:

PSMId   score   q-value peptide Label   proteinIds
target_0_6395_3_1       0.999988        0.385129        K.IDSAAETHADAPVVDASPAEDQASEVTEAPHVESAK.S        1       F52H3.7a
target_0_8821_3_1       0.999975        0.385129        R.LHCTAQPMPDGLADDIEGGTVNAR.D    1       F25H5.4
target_0_10580_2_1      0.999968        0.385129        K.LANALEPGAVEVAAAEENADAAAQK.E   1       C10G11.7
target_0_16058_2_1      0.999959        0.385129        K.TVDVISDTGTSFLGGPQSVVDGLAK.A   1       F21F8.7
target_0_17089_2_1      0.999956        0.385129        K.HTDAVAELTDQLDQLNK.A   1       F11C3.3
target_0_16837_2_1      0.999955        0.385129        K.HTDAVAELTDQLDQLNK.A   1       F11C3.3
target_0_27067_3_1      0.999952        0.385129        R.FQSSAVMALQEAAEAYLVGLFEDTNLCAIHAK.R    1       ZK131.3
target_0_15165_2_1      0.999945        0.385129        R.NLQIAQGTPGGLITYGAIDTVNCAK.Q   1       Y39B6A.20
target_0_17414_2_1      0.999943        0.385129        K.VTLEDQIVQTNPVLEAFGNAK.T       1       F11C3.3

Further analysis options are discussed in the next section.

Q-value analysis plots

Resulting PSM identifications vs q-values may be easily plotted and compared against other methods. Assuming test.pin has feature XCorr specified in its header, the following may be run on the command line to plot target-decoy competition (TDC) results for both recalibrated ProteoTorch scores and the uncalibrated XCorr scores:

proteoTorchPlot --output test.pdf --maxq 0.1  \
    --tdc --dataset test.pin \
    "ProteoTorch DNN":"score":output_dir/output.txt \
    "XCorr":'XCorr':test.pin

In the above, --output specifies the resulting plot file name (and format), --maxq specifies the maximum q-value threshold to plot, --tdc runs TDC for all methods, and tab-delimited PSM files are specified by the triple method:header field:PSM file. Note that, when TDC is specified, the original pin file must be specified using --dataset to properly calculate the experimental mass, scan number identifiers for each PSM.

Further details and examples are discussed on the plotting page.