Hi. I run Comet and Peptideprophet on two public datasets with TDC with UniProt. I calculated the q-value in Python based on fval distribution and filtered data with a threshold 1%. Like that: #df - PeptideProphet outputdf = df.sort_values(by='fval', ascending=False).reset_index(drop=True) # Calculate cumulative counts of targets and decoys df['cum_targets'] = (df['database'] == 'T').cumsum() df['cum_decoys'] = (df['database'] == 'D').cumsum() # Calculate FDR df['FDR'] = df['cum_decoys'] / df['cum_targets'] # cumulative minimum from bottom to top df['q-value'] = df['FDR'][::-1].cummin()[::-1]
Lower you see the proportion of PSMs annotated as targets and decoys which passed the value threshold or not. One of the datasets (PXD03594) has a very low number of identifications. It also has a wide distribution of decoys (on the graph the raw files are plotted together). Could anyone suggest what could have happened here? I used default parameters, just changed peptide length from 7 to 30 aa, and peptide mass range 500.0-6000.0, and also enabled Methioning clipping. Thanks! [image: Capture9.PNG][image: Capture8.PNG] [image: Capture6.PNG][image: Capture7.PNG] -- You received this message because you are subscribed to the Google Groups "spctools-discuss" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/spctools-discuss/e8201f2e-0fd1-4c36-ac34-dbeada186d13n%40googlegroups.com.
