Python 2.7.

1. mgfextract_notebook.ipynb • It extracts specific subsets from an MGF file by matching the original MGF file from against the ground truth (Excel file); • Uses the method (write_title) from mgf_extract.py to create the required MGF files.

2. mgf_extract.py • It splits the original MGF file into multiple folders (one spectrum per folder) and then chooses the required spectra (based on the title given to each folder which contains information such as the protein id and the protein sequence), deletes the unwanted folder and merges the rest of them into one folder.

3. data_pre_analysis.ipynb • It was used for performing the analysis for the first part of the project. • It uses mgf_parser.py for parsing the mgf file. • It uses mgf_analysis.py for assigning and computing all the required values to each spectrum. • Creates the histogram of fragment differences found in both phosphorylated and non-PTM spectra. Takes some time to compute (~7 min), due to v. large number of differences. • Calculates the percentage of spectra matching with fragments matching the calculated values and the percentages of y and b ions for each matching spectrum. • Creates the graph for determining the mass tolerance threshold.

4. mgf_parser.py • The main code for this was developed by PhD student Joe Wandy. I’ve only adapted it for parsing a slightly modified format of mgf file (with a few extra lines added to it for 1)protein id; 2)mh+; 3)protein sequence)

5. mgf_analysis.py • Assigns and computes all the required values to each spectrum as shown in Figure 1. • It uses chem_utils.py and the ‘Document’ and ‘Peak’ objects from ms_objects.py.

6. chem_utils.py • It creates a dictionary with the masses of all amino acids and calculates the masses of y and b ions when given a protein sequence.

7. ms_objects.py • It is only used for creating useful objects to be used in other scripts.

8. new_lda_run_phos_eta_1.ipynb (all three: phos_eta_1, phos_eta_01, acetyl) • It was used for performing the analysis for the second part of the project. • It uses feature_extractor.py (feature_extractor_acetyl.py for acetylated spectra) for preparing the dictionary (‘corpus’) which is going to be used for the LDA run. • It creates the PCA plot • It splits the data to be used for the classifier and performs the evaluation for each classifier.

9. feature_extractor.py • It is used for grouping very similar differences (within a certain mass tolerance) from the list of differences of all spectra. The intensity assigned to each group within a spectrum is the average of all the intensities assigned to the differences in the respective group (each difference is assigned the lower intensity between the two fragments). • A word is assigned to each group base on the mean value of the differences in the group, i.e diff_mean-value

. Each MS1 peak is identified by its Peak ID (extracted from the ground truth), m/z, rt, intensity and charge and MS2 values (extracted from the MGF file). For each MS1 spectra tandem MS values were matched with theoretical values of fragments and the matching percentage was calculated