Feature/sage

spectrum_io/search_result/__init__.py

@@ -2,3 +2,4 @@
 from .mascot import Mascot
 from .maxquant import MaxQuant
 from .msfragger import MSFragger
+from .sage import Sage

spectrum_io/search_result/sage.py

@@ -0,0 +1,78 @@
+import logging
+from pathlib import Path
+from typing import Union
+
+import pandas as pd
+import spectrum_fundamentals.constants as c
+from spectrum_fundamentals.mod_string import sage_to_internal
+
+from .search_results import SearchResults, filter_valid_prosit_sequences
+
+logger = logging.getLogger(__name__)
+
+
+class Sage(SearchResults):
+    """Handle search results from Sage."""
+
+    @staticmethod
+    def read_result(path: Union[str, Path], tmt_labeled: str = "") -> pd.DataFrame:
+        """
+        Function to read a msms tsv and perform some basic formatting.
+
+        :param path: path to msms.tsv to read
+        :param tmt_labeled: tmt label as str
+        :return: pd.DataFrame with the formatted data
+        """
+        logger.info("Reading msms.tsv file")
+        df = pd.read_csv(
+            path,
+            usecols=["filename", "scannr", "peptide", "charge", "hyperscore", "calcmass", "proteins"],
+            sep="\t",
+        )
+        logger.info("Finished reading msms.tsv file")
+
+        # Standardize column names
+        df.columns = df.columns.str.upper()
+        df.columns = df.columns.str.replace(" ", "_")
+
+        df = Sage.update_columns_for_prosit(df, tmt_labeled)
+        return filter_valid_prosit_sequences(df)
+
+    @staticmethod
+    def update_columns_for_prosit(df: pd.DataFrame, tmt_labeled: str) -> pd.DataFrame:
+        """
+        Update columns of df to work with Prosit.
+
+        :param df: df to modify
+        :param tmt_labeled: True if tmt labeled, ignored
+        :return: modified df as pd.DataFrame
+        """
+        df = df.rename(
+            columns={
+                "FILENAME": "RAW_FILE",
+                "SCANNR": "SCAN_NUMBER",
+                "PEPTIDE": "MODIFIED_SEQUENCE",
+                "CHARGE": "PRECURSOR_CHARGE",
+            }
+        )
+
+        # removing .mzML
+        df["RAW_FILE"] = df["RAW_FILE"].str.replace(".mzML", "", regex=True)
+        # extracting only the scan number
+        df["SCAN_NUMBER"] = [int(x.rsplit("=", 1)[-1]) for x in df["SCAN_NUMBER"]]
+        # creating a column of decoys and targets
+        df["REVERSE"] = df["PROTEINS"].str.startswith("rev_")
+        # removing modification to create the unmodified sequences
+        df["SEQUENCE"] = df["MODIFIED_SEQUENCE"].str.replace(r"\[.*?\]", "", regex=True)
+        # length of the peptide
+        df["PEPTIDE_LENGTH"] = df["SEQUENCE"].str.len()
+        # mass of the peptide
+        df["MASS"] = df["CALCMASS"]
+        # score of the peptide
+        df["SCORE"] = df["HYPERSCORE"]
+        # converting proforma to unimode
+        print(df)
+        df["MODIFIED_SEQUENCE"] = sage_to_internal(df["MODIFIED_SEQUENCE"])
+
+        print(df.columns)
+        return df

tests/unit_tests/data/sage_output.tsv

@@ -0,0 +1,18 @@
+peptide	proteins	num_proteins	filename	scannr	rank	label	expmass	calcmass	charge	peptide_len	missed_cleavages	isotope_error	precursor_ppm	fragment_ppm	hyperscore	delta_next	delta_best	rt	aligned_rt	predicted_rt	delta_rt_model	matched_peaks	longest_b	longest_y	longest_y_pct	matched_intensity_pct	scored_candidates	poisson	sage_discriminant_score	posterior_error	spectrum_q	peptide_q	protein_q	ms1_intensity	ms2_intensity
+[+229.1629]-HLDGGAEQSLLFVAGM[+15.9949]R	rev_sp|P26006-1|ITA3_HUMAN;rev_sp|P26006|ITA3_HUMAN	2	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=50989	1	-1	2199.172	2045.0568	2	17	0	0.0	72623.49	7.826053	18.084579770792818	0.20642824189008735	0.0	113.94822	0.9495685	0.0	0.9495685	4	2	1	0.05882353	1.318411	84602	-2.0562283019710534	-0.37137848	-0.35820845	0.30232558	0.5880901	0.5815684	926232.0	7361.109
+[+229.1629]-GRFVEPLSNVQEEWNQK[+229.1629]	rev_sp|Q9H1A3-2|METL9_HUMAN;rev_sp|Q9H1A3|METL9_HUMAN;rev_tr|H3BN86|H3BN86_HUMAN;rev_tr|Q8TD49|Q8TD49_HUMAN	4	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=37495	1	-1	2671.5417	2517.338	2	17	1	0.0	59436.277	4.4063077	21.44256145543081	0.24493444146309074	0.0	83.519325	0.6959944	0.0	0.6959944	4	1	2	0.11764706	3.139837	130027	-2.014281373814444	-0.37227735	-0.35820845	0.30232558	0.5880901	0.5815684	2674831.2	42405.516
+[+229.1629]-LTVEC[+57.0214]MPTIASDDLPVGTLQESEVSM[+15.9949]TGPG	rev_tr|C9JVX2|C9JVX2_HUMAN	1	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=45450	1	-1	3533.4688	3378.6082	3	30	0	0.0	44808.7	2.583788	18.628243636678995	0.4911733270242351	0.0	102.02221	0.8501851	0.0	0.8501851	4	1	2	0.06666667	1.6194082	80638	-2.046675752396668	-0.3724502	-0.35820845	0.30232558	0.5880901	0.5815684	643726.75	8588.239
+[+229.1629]-VNM[+15.9949]RTSSSIQNEDEATSMELIAPGP	sp|Q9Y5Y9|SCNAA_HUMAN;tr|A0A2R8Y6J6|A0A2R8Y6J6_HUMAN;tr|A0A590UJM0|A0A590UJM0_HUMAN	3	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=45493	1	1	3074.5525	2921.3948	3	25	1	0.0	51087.08	5.391072	19.979716393086438	0.2293833747943026	0.0	102.12942	0.85107845	0.0	0.85107845	5	1	2	0.08	1.4697478	154744	-2.71207222963538	-0.3737817	-0.35820845	0.30232558	0.5880901	0.5815684	1085421.8	8865.469
+[+229.1629]-VGEQEAPHEGGHPGSDSARASMADWLR	sp|Q9H093|NUAK2_HUMAN	1	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=41946	1	1	3229.3804	3075.4417	3	27	1	0.0	48832.06	2.6081188	20.936078174830033	0.05364075290029291	0.0	93.63916	0.7803263	0.0	0.7803263	4	1	1	0.037037037	6.790568	109802	-2.034297610829216	-0.37500644	-0.35820845	0.30232558	0.5880901	0.5815684	604286.9	36628.875
+[+229.1629]-EM[+15.9949]VSPTDSC[+57.0214]VRVSVRDLPQFHVSVVDM[+15.9949]DR	rev_sp|Q9HCN3|PGAP6_HUMAN;rev_tr|K4DI83|K4DI83_HUMAN	2	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=39750	1	-1	3924.602	3620.7583	3	29	2	0.0	80537.9	4.2269964	17.954540767022678	0.13827490088740646	0.0	88.40008	0.73666734	0.0	0.73666734	4	1	2	0.06896552	2.1663432	50127	-2.0296029706779146	-0.375757	-0.35820845	0.30232558	0.5880901	0.5815684	1711349.0	6906.8516
+[+229.1629]-K[+229.1629]M[+15.9949]EEDIYTNLSK[+229.1629]METVLGQSMSSLPLSYR	sp|Q8WXH0-2|SYNE2_HUMAN;sp|Q8WXH0|SYNE2_HUMAN;tr|A0A0A0MRE3|A0A0A0MRE3_HUMAN;tr|G3V5X4|G3V5X4_HUMAN	4	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=45441	2	1	4205.6133	4053.1077	4	29	2	0.0	36932.02	7.9742827	19.962357128713123	0.0	0.0	101.99851	0.8499876	0.0	0.8499876	4	1	2	0.06896552	2.6182032	65126	-1.9825140292603154	-0.37760013	-0.35820845	0.30232558	0.5880901	0.5815684	1558514.0	15372.994
+[+229.1629]-TC[+57.0214]SK[+229.1629]SQGSWGNREIVIIDTPDMFSWK[+229.1629]	sp|Q9UG22|GIMA2_HUMAN	1	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=36758	1	1	3881.789	3728.9263	3	26	2	0.0	40170.414	8.588704	18.776753308950084	2.853361183504054	0.0	81.83095	0.6819246	0.0	0.6819246	4	1	1	0.03846154	2.170038	54463	-2.02115128698497	-0.3785715	-0.35820845	0.30232558	0.5880901	0.5815684	1264436.8	13284.256
+[+229.1629]-M[+15.9949]SLGRAAPSAPGR	rev_sp|P51693-2|APLP1_HUMAN;rev_sp|P51693|APLP1_HUMAN;rev_tr|K7EMS1|K7EMS1_HUMAN	3	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=49100	1	-1	1669.267	1514.819	2	13	1	0.0	97012.45	4.073574	20.66450790410668	0.2021918361478079	0.0	109.23304	0.91027534	0.0	0.91027534	5	1	3	0.23076923	1.8854505	176106	-2.684809584132347	-0.37956935	-0.35820845	0.30232558	0.5880901	0.5815684	611105.25	12195.32
+[+229.1629]-C[+57.0214]LIQM[+15.9949]GAAVEAK[+229.1629]AYNGNTALHVAASLQYR	tr|H7C5S1|H7C5S1_HUMAN	1	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=37376	1	1	3897.3843	3593.885	3	29	1	0.0	81027.46	5.0039053	21.88178474607965	0.07848479239995854	0.0	83.277466	0.6939789	0.0	0.6939789	4	1	2	0.06896552	2.1495197	58331	-2.045800246831915	-0.37965888	-0.35820845	0.30232558	0.5880901	0.5815684	18796730.0	52520.473
+[+229.1629]-M[+15.9949]EESLNIVK[+229.1629]YTAFLYNDQLIWSGLEQDDMR	sp|P86790|CCZ1B_HUMAN;sp|P86791|CCZ1_HUMAN	2	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=41296	1	1	4398.7207	4095.037	3	30	1	0.0	71507.48	3.7571762	22.91659667770412	2.159702635767953	0.0	92.108765	0.76757306	0.0	0.76757306	4	2	2	0.06666667	18.10546	11995	-2.0440490580574484	-0.38045537	-0.35820845	0.30232558	0.5880901	0.5815684	1328658.9	71828.805
+[+229.1629]-LNVEGTERGSC[+57.0214]GRK[+229.1629]	sp|O75078-2|ADA11_HUMAN;sp|O75078|ADA11_HUMAN;tr|B4DKD2|B4DKD2_HUMAN;tr|K7EKA8|K7EKA8_HUMAN	4	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=40784	2	1	2174.1533	2020.0887	3	14	2	0.0	73464.8	5.374492	21.842498887625368	0.6170763228642393	0.0	90.874275	0.75728565	0.0	0.75728565	5	3	2	0.14285715	2.8490458	237683	-2.6522210241481585	-0.3812077	-0.35820845	0.30232558	0.5880901	0.5815684	1290087.5	21134.262
+[+229.1629]-C[+57.0214]NRGWTALHESVSR	sp|Q96Q27-1|ASB2_HUMAN;sp|Q96Q27|ASB2_HUMAN;tr|G3V2Z2|G3V2Z2_HUMAN;tr|G3V4B2|G3V4B2_HUMAN	4	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=51002	2	1	2055.9688	1900.9526	2	14	1	0.0	78351.875	5.277066	17.823081883875247	0.012034867562086049	0.0	113.98048	0.9498373	0.0	0.9498373	4	1	1	0.071428575	1.2503061	115129	-2.0307522097044646	-0.3828116	-0.35820845	0.30232558	0.5880901	0.5815684	977832.06	5506.2334
+[+229.1629]-AEVDNQMHVVDK[+229.1629]NPVSLVSK[+229.1629]TR	rev_sp|O75151|PHF2_HUMAN	1	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=40715	2	-1	3307.599	3152.7583	3	22	2	0.0	47935.68	2.6617033	19.971965322167343	1.5655510246394897	0.0	90.70834	0.7559029	0.0	0.7559029	4	2	1	0.045454547	2.7450855	107384	-2.0235551137874417	-0.38507006	-0.35820845	0.30232558	0.5880901	0.5815684	923512.1	14331.008
+[+229.1629]-YLLSLEEERPALMDDR	sp|Q86TB9-2|PATL1_HUMAN;sp|Q86TB9-4|PATL1_HUMAN;sp|Q86TB9|PATL1_HUMAN	3	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=37506	1	1	2331.2417	2178.1194	2	16	0	0.0	67913.086	2.5317562	19.997613788592076	0.39086524401945155	0.0	83.542175	0.6961848	0.0	0.6961848	4	1	2	0.125	4.0383697	80445	-2.062325498085077	-0.38619247	-0.35820845	0.30232558	0.5880901	0.5815684	3893642.5	14758.541
+[+229.1629]-EGRGAGSQSPPRGR	sp|Q6ZSN1|YI023_HUMAN	1	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=45427	1	1	1942.9047	1639.8704	2	14	2	0.0	169161.78	4.145797	19.229664162247772	0.4602391019734071	0.0	101.96355	0.8496962	0.0	0.8496962	4	1	1	0.071428575	4.303446	136956	-2.026177700336816	-0.3904131	-0.35820845	0.30232558	0.5880901	0.5815684	993124.44	14842.607
+[+229.1629]-TASASRRSAR	sp|P08729|K2C7_HUMAN	1	GN20170722_SK_HLA_G0103_R1_01.mzML	controllerType=0 controllerNumber=1 scan=30700	1	1	1444.0938	1290.7319	2	10	2	0.0	112154.734	2.7423766	19.376349012983994	0.42494292662552624	0.0	67.56542	0.5630452	0.0	0.5630452	4	1	2	0.2	4.1968794	97885	-2.029718921235703	-0.3916049	-0.35820845	0.30232558	0.5880901	0.5815684	266738.5	11889.021

tests/unit_tests/data/sage_output_internal.csv

@@ -0,0 +1,16 @@
+,MODIFIED_SEQUENCE,PROTEINS,RAW_FILE,SCAN_NUMBER,CALCMASS,PRECURSOR_CHARGE,HYPERSCORE,REVERSE,SEQUENCE,PEPTIDE_LENGTH,MASS,SCORE
+0,[UNIMOD:737]-HLDGGAEQSLLFVAGM[UNIMOD:35]R,rev_sp|P26006-1|ITA3_HUMAN;rev_sp|P26006|ITA3_HUMAN,GN20170722_SK_HLA_G0103_R1_01,50989,2045.0568,2,18.084579770792818,True,-HLDGGAEQSLLFVAGMR,18,2045.0568,18.084579770792818
+1,[UNIMOD:737]-GRFVEPLSNVQEEWNQK[UNIMOD:737],rev_sp|Q9H1A3-2|METL9_HUMAN;rev_sp|Q9H1A3|METL9_HUMAN;rev_tr|H3BN86|H3BN86_HUMAN;rev_tr|Q8TD49|Q8TD49_HUMAN,GN20170722_SK_HLA_G0103_R1_01,37495,2517.338,2,21.44256145543081,True,-GRFVEPLSNVQEEWNQK,18,2517.338,21.44256145543081
+3,[UNIMOD:737]-VNM[UNIMOD:35]RTSSSIQNEDEATSMELIAPGP,sp|Q9Y5Y9|SCNAA_HUMAN;tr|A0A2R8Y6J6|A0A2R8Y6J6_HUMAN;tr|A0A590UJM0|A0A590UJM0_HUMAN,GN20170722_SK_HLA_G0103_R1_01,45493,2921.3948,3,19.97971639308644,False,-VNMRTSSSIQNEDEATSMELIAPGP,26,2921.3948,19.97971639308644
+4,[UNIMOD:737]-VGEQEAPHEGGHPGSDSARASMADWLR,sp|Q9H093|NUAK2_HUMAN,GN20170722_SK_HLA_G0103_R1_01,41946,3075.4417,3,20.936078174830037,False,-VGEQEAPHEGGHPGSDSARASMADWLR,28,3075.4417,20.936078174830037
+5,[UNIMOD:737]-EM[UNIMOD:35]VSPTDSC[UNIMOD:4]VRVSVRDLPQFHVSVVDM[UNIMOD:35]DR,rev_sp|Q9HCN3|PGAP6_HUMAN;rev_tr|K4DI83|K4DI83_HUMAN,GN20170722_SK_HLA_G0103_R1_01,39750,3620.7583,3,17.954540767022678,True,-EMVSPTDSCVRVSVRDLPQFHVSVVDMDR,30,3620.7583,17.954540767022678
+6,[UNIMOD:737]-K[UNIMOD:737]M[UNIMOD:35]EEDIYTNLSK[UNIMOD:737]METVLGQSMSSLPLSYR,sp|Q8WXH0-2|SYNE2_HUMAN;sp|Q8WXH0|SYNE2_HUMAN;tr|A0A0A0MRE3|A0A0A0MRE3_HUMAN;tr|G3V5X4|G3V5X4_HUMAN,GN20170722_SK_HLA_G0103_R1_01,45441,4053.1077,4,19.962357128713123,False,-KMEEDIYTNLSKMETVLGQSMSSLPLSYR,30,4053.1077,19.962357128713123
+7,[UNIMOD:737]-TC[UNIMOD:4]SK[UNIMOD:737]SQGSWGNREIVIIDTPDMFSWK[UNIMOD:737],sp|Q9UG22|GIMA2_HUMAN,GN20170722_SK_HLA_G0103_R1_01,36758,3728.9263,3,18.776753308950084,False,-TCSKSQGSWGNREIVIIDTPDMFSWK,27,3728.9263,18.776753308950084
+8,[UNIMOD:737]-M[UNIMOD:35]SLGRAAPSAPGR,rev_sp|P51693-2|APLP1_HUMAN;rev_sp|P51693|APLP1_HUMAN;rev_tr|K7EMS1|K7EMS1_HUMAN,GN20170722_SK_HLA_G0103_R1_01,49100,1514.819,2,20.66450790410668,True,-MSLGRAAPSAPGR,14,1514.819,20.66450790410668
+9,[UNIMOD:737]-C[UNIMOD:4]LIQM[UNIMOD:35]GAAVEAK[UNIMOD:737]AYNGNTALHVAASLQYR,tr|H7C5S1|H7C5S1_HUMAN,GN20170722_SK_HLA_G0103_R1_01,37376,3593.885,3,21.88178474607965,False,-CLIQMGAAVEAKAYNGNTALHVAASLQYR,30,3593.885,21.88178474607965
+11,[UNIMOD:737]-LNVEGTERGSC[UNIMOD:4]GRK[UNIMOD:737],sp|O75078-2|ADA11_HUMAN;sp|O75078|ADA11_HUMAN;tr|B4DKD2|B4DKD2_HUMAN;tr|K7EKA8|K7EKA8_HUMAN,GN20170722_SK_HLA_G0103_R1_01,40784,2020.0887,3,21.842498887625368,False,-LNVEGTERGSCGRK,15,2020.0887,21.842498887625368
+12,[UNIMOD:737]-C[UNIMOD:4]NRGWTALHESVSR,sp|Q96Q27-1|ASB2_HUMAN;sp|Q96Q27|ASB2_HUMAN;tr|G3V2Z2|G3V2Z2_HUMAN;tr|G3V4B2|G3V4B2_HUMAN,GN20170722_SK_HLA_G0103_R1_01,51002,1900.9526,2,17.823081883875247,False,-CNRGWTALHESVSR,15,1900.9526,17.823081883875247
+13,[UNIMOD:737]-AEVDNQMHVVDK[UNIMOD:737]NPVSLVSK[UNIMOD:737]TR,rev_sp|O75151|PHF2_HUMAN,GN20170722_SK_HLA_G0103_R1_01,40715,3152.7583,3,19.971965322167343,True,-AEVDNQMHVVDKNPVSLVSKTR,23,3152.7583,19.971965322167343
+14,[UNIMOD:737]-YLLSLEEERPALMDDR,sp|Q86TB9-2|PATL1_HUMAN;sp|Q86TB9-4|PATL1_HUMAN;sp|Q86TB9|PATL1_HUMAN,GN20170722_SK_HLA_G0103_R1_01,37506,2178.1194,2,19.99761378859208,False,-YLLSLEEERPALMDDR,17,2178.1194,19.99761378859208
+15,[UNIMOD:737]-EGRGAGSQSPPRGR,sp|Q6ZSN1|YI023_HUMAN,GN20170722_SK_HLA_G0103_R1_01,45427,1639.8704,2,19.22966416224777,False,-EGRGAGSQSPPRGR,15,1639.8704,19.22966416224777
+16,[UNIMOD:737]-TASASRRSAR,sp|P08729|K2C7_HUMAN,GN20170722_SK_HLA_G0103_R1_01,30700,1290.7319,2,19.376349012983997,False,-TASASRRSAR,11,1290.7319,19.376349012983997

tests/unit_tests/test_sage.py

@@ -0,0 +1,20 @@
+import unittest
+from pathlib import Path
+
+import pandas as pd
+
+from spectrum_io.search_result import Sage
+
+
+class TestSage(unittest.TestCase):
+    """Test vlass to check Sage search result processing."""
+
+    def test_read_sage(self):
+        """Test function for reading sage results and transforming to Prosit format."""
+        sage_output_path = Path(__file__).parent / "data" / "sage_output.tsv"
+        expected_sage_internal_path = Path(__file__).parent / "data" / "sage_output_internal.csv"
+
+        internal_search_results_df = Sage.read_result(sage_output_path)
+        expected_df = pd.read_csv(expected_sage_internal_path, index_col=0)
+
+        pd.testing.assert_frame_equal(internal_search_results_df, expected_df)