generic assay binary data and categorical data validator

core/src/main/scripts/importer/allowed_data_types.txt

@@ -59,4 +59,6 @@ PROTEIN_LEVEL	Z-SCORE	protein_quantification_zscores
 GENESET_SCORE	GSVA-SCORE	gsva_scores
 GENESET_SCORE	P-VALUE	gsva_pvalues
 GENERIC_ASSAY	LIMIT-VALUE	*
+GENERIC_ASSAY	BINARY	*
+GENERIC_ASSAY	CATEGORICAL	*
 STRUCTURAL_VARIANT	SV	structural_variants

core/src/main/scripts/importer/cbioportal_common.py

@@ -60,7 +60,9 @@ class MetaFileTypes(object):
     GENE_PANEL_MATRIX = 'meta_gene_panel_matrix'
     GSVA_SCORES = 'meta_gsva_scores'
     GSVA_PVALUES = 'meta_gsva_pvalues'
-    GENERIC_ASSAY = 'meta_generic_assay'
+    GENERIC_ASSAY_CONTINUOUS = 'meta_generic_assay_continuous'
+    GENERIC_ASSAY_BINARY = 'meta_generic_assay_binary'
+    GENERIC_ASSAY_CATEGORICAL = 'meta_generic_assay_categorical'
     STRUCTURAL_VARIANT = 'meta_structural_variants'
     SAMPLE_RESOURCES = 'meta_resource_sample'
     PATIENT_RESOURCES = 'meta_resource_patient'
@@ -257,7 +259,7 @@ class MetaFileTypes(object):
         'show_profile_in_analysis_tab': True,
         'geneset_def_version': True
     },
-    MetaFileTypes.GENERIC_ASSAY: {
+    MetaFileTypes.GENERIC_ASSAY_CONTINUOUS: {
         'cancer_study_identifier': True,
         'genetic_alteration_type': True,
         'generic_assay_type': True,
@@ -271,6 +273,30 @@ class MetaFileTypes(object):
         'pivot_threshold_value': False,
         'value_sort_order': False
     },
+    MetaFileTypes.GENERIC_ASSAY_BINARY: {
+        'cancer_study_identifier': True,
+        'genetic_alteration_type': True,
+        'generic_assay_type': True,
+        'datatype': True,
+        'stable_id': True,
+        'profile_name': True,
+        'profile_description': True,
+        'data_filename': True,
+        'show_profile_in_analysis_tab': True,
+        'generic_entity_meta_properties': False
+    },
+    MetaFileTypes.GENERIC_ASSAY_CATEGORICAL: {
+        'cancer_study_identifier': True,
+        'genetic_alteration_type': True,
+        'generic_assay_type': True,
+        'datatype': True,
+        'stable_id': True,
+        'profile_name': True,
+        'profile_description': True,
+        'data_filename': True,
+        'show_profile_in_analysis_tab': True,
+        'generic_entity_meta_properties': False
+    },
     MetaFileTypes.STRUCTURAL_VARIANT: {
         'cancer_study_identifier': True,
         'genetic_alteration_type': True,
@@ -325,7 +351,9 @@ class MetaFileTypes(object):
     MetaFileTypes.GENE_PANEL_MATRIX: "org.mskcc.cbio.portal.scripts.ImportGenePanelProfileMap",
     MetaFileTypes.GSVA_SCORES: "org.mskcc.cbio.portal.scripts.ImportProfileData",
     MetaFileTypes.GSVA_PVALUES: "org.mskcc.cbio.portal.scripts.ImportProfileData",
-    MetaFileTypes.GENERIC_ASSAY: "org.mskcc.cbio.portal.scripts.ImportProfileData",
+    MetaFileTypes.GENERIC_ASSAY_CONTINUOUS: "org.mskcc.cbio.portal.scripts.ImportProfileData",
+    MetaFileTypes.GENERIC_ASSAY_BINARY: "org.mskcc.cbio.portal.scripts.ImportProfileData",
+    MetaFileTypes.GENERIC_ASSAY_CATEGORICAL: "org.mskcc.cbio.portal.scripts.ImportProfileData",
     MetaFileTypes.STRUCTURAL_VARIANT: "org.mskcc.cbio.portal.scripts.ImportProfileData",
     MetaFileTypes.SAMPLE_RESOURCES: "org.mskcc.cbio.portal.scripts.ImportResourceData",
     MetaFileTypes.PATIENT_RESOURCES: "org.mskcc.cbio.portal.scripts.ImportResourceData",
@@ -594,7 +622,9 @@ def get_meta_file_type(meta_dictionary, logger, filename):
         ("MUTSIG", "Q-VALUE"): MetaFileTypes.MUTATION_SIGNIFICANCE,
         ("GENESET_SCORE", "GSVA-SCORE"): MetaFileTypes.GSVA_SCORES,
         ("GENESET_SCORE", "P-VALUE"): MetaFileTypes.GSVA_PVALUES,
-        ("GENERIC_ASSAY", "LIMIT-VALUE"): MetaFileTypes.GENERIC_ASSAY
+        ("GENERIC_ASSAY", "LIMIT-VALUE"): MetaFileTypes.GENERIC_ASSAY_CONTINUOUS,
+        ("GENERIC_ASSAY", "BINARY"): MetaFileTypes.GENERIC_ASSAY_BINARY,
+        ("GENERIC_ASSAY", "CATEGORICAL"): MetaFileTypes.GENERIC_ASSAY_CATEGORICAL
     }
     result = None
     if 'genetic_alteration_type' in meta_dictionary and 'datatype' in meta_dictionary:

core/src/main/scripts/importer/validateData.py

@@ -108,7 +108,9 @@
     cbioportal_common.MetaFileTypes.GENE_PANEL_MATRIX:'GenePanelMatrixValidator',
     cbioportal_common.MetaFileTypes.GSVA_SCORES:'GsvaScoreValidator',
     cbioportal_common.MetaFileTypes.GSVA_PVALUES:'GsvaPvalueValidator',
-    cbioportal_common.MetaFileTypes.GENERIC_ASSAY:'GenericAssayValidator',
+    cbioportal_common.MetaFileTypes.GENERIC_ASSAY_CONTINUOUS:'GenericAssayContinuousValidator',
+    cbioportal_common.MetaFileTypes.GENERIC_ASSAY_BINARY:'GenericAssayBinaryValidator',
+    cbioportal_common.MetaFileTypes.GENERIC_ASSAY_CATEGORICAL:'GenericAssayCategoricalValidator',
     cbioportal_common.MetaFileTypes.STRUCTURAL_VARIANT:'StructuralVariantValidator',
     cbioportal_common.MetaFileTypes.SAMPLE_RESOURCES:'SampleResourceValidator',
     cbioportal_common.MetaFileTypes.PATIENT_RESOURCES:'PatientResourceValidator',
@@ -4334,6 +4336,7 @@ def __init__(self, *args, **kwargs):
     REQUIRED_HEADERS = ['ENTITY_STABLE_ID']
     OPTIONAL_HEADERS = []
     UNIQUE_COLUMNS = ['ENTITY_STABLE_ID']
+    NULL_VALUES = ["NA"]
 
     def parseFeatureColumns(self, nonsample_col_vals):
         """Check the IDs in the first column."""
@@ -4343,21 +4346,15 @@ def parseFeatureColumns(self, nonsample_col_vals):
                               extra={'line_number': self.line_number,
                                      'column_number': 1,
                                      'cause': nonsample_col_vals[0]})
-            return None
-
-    def checkValue(self, value, col_index):
-        """Check a value in a sample column."""
-        stripped_value = value.strip()
-        if stripped_value not in self.NULL_VALUES and not self.checkFloat(stripped_value):
-            self.logger.error("Value is neither a real number nor " + ', '.join(self.NULL_VALUES),
-                              extra={'line_number': self.line_number,
-                                     'column_number': col_index + 1,
-                                     'cause': value})
+        return value
 
-class GenericAssayValidator(GenericAssayWiseFileValidator):
+class GenericAssayContinuousValidator(GenericAssayWiseFileValidator):
 
-    """ Validator for files containing generic assay values.
+    """ Validator for files containing generic assay limit continuous values.
     """
+    def __init__(self, *args, **kwargs):
+        """Initialize the instance attributes of the data file validator."""
+        super(GenericAssayContinuousValidator, self).__init__(*args, **kwargs)
 
     # (1) Natural positive number (not 0)
     # (2) Number may be prefixed by ">" or "<"; f.i. ">n" means that the real value lies beyond value n.
@@ -4367,56 +4364,113 @@ class GenericAssayValidator(GenericAssayWiseFileValidator):
     # (1) Cell contains a value without decimals and is not prefixed by ">"; value appears to be truncated but lacks ">" truncation indicator
     def checkValue(self, value, col_index):
         """Check a value in a sample column."""
+        stripped_value = value.strip()
+        # if the value is prefixed with '>' or '<' remove this prefix
+        # prior to evaluation of the numeric value
+        hasTruncSymbol = re.match("^[><]", stripped_value)
+        stripped_value = re.sub(r"^[><]\s*","", stripped_value)
+
+        # do not check null values
+        # 'NA' is an allowed value. No further validations apply.
+        if stripped_value in self.NULL_VALUES:
+            return
 
         # value is not defined (empty cell)
-        stripped_value = value.strip()
-        if stripped_value == "":
-            self.logger.error("Cell is empty. A response value value is expected. Use 'NA' to indicate missing values.",
+        if len(stripped_value) == 0:
+            self.logger.error("Cell is empty. A value is expected. Use 'NA' to indicate missing values.",
                 extra={'line_number': self.line_number,
                 'column_number': col_index + 1,
                 'cause': value})
             return
 
-        # 'NA' is an allowed value. No further validations apply.
-        if stripped_value == 'NA':
-            return
-
-        # if the value is prefixed with '>' or '<' remove this prefix
-        # prior to evaluation of the numeric value
-        hasTruncSymbol = re.match("^[><]", stripped_value)
-        stripped_value = re.sub(r"^[><]\s*","", stripped_value)
-
         try:
             numeric_value = float(stripped_value)
         except ValueError:
-            self.logger.error("Value cannot be interpreted as a floating point number and is not valid response value.",
+            self.logger.error("Value cannot be interpreted as a floating point number and is not valid value.",
                 extra={'line_number': self.line_number,
                 'column_number': col_index + 1,
                 'cause': value})
             return
 
         if math.isnan(numeric_value):
-            self.logger.error("Value is NaN, therefore, not a valid response value.",
+            self.logger.error("Value is NaN, therefore, not a valid value.",
                 extra={'line_number': self.line_number,
                 'column_number': col_index + 1,
                 'cause': value})
             return
 
         if math.isinf(numeric_value):
-            self.logger.error("Value is infinite and, therefore, not a valid response value.",
+            self.logger.error("Value is infinite and, therefore, not a valid value.",
                 extra={'line_number': self.line_number,
                 'column_number': col_index + 1,
                 'cause': value})
             return
 
         if numeric_value % 1 == 0 and not hasTruncSymbol:
-            self.logger.warning("Value has no decimals and may represent an invalid response value.",
+            self.logger.warning("Value has no decimals and may represent an invalid value.",
                 extra={'line_number': self.line_number,
                 'column_number': col_index + 1,
                 'cause': value})
 
         return
 
+class GenericAssayCategoricalValidator(GenericAssayWiseFileValidator):
+
+    """ Validator for files containing generic assay categorical values.
+    """
+    def __init__(self, *args, **kwargs):
+        """Initialize the instance attributes of the data file validator."""
+        super(GenericAssayCategoricalValidator, self).__init__(*args, **kwargs)
+
+    # (1) non-empty string
+    # (2) NA cell value is allowed; means value was not tested on a sample
+
+    def checkValue(self, value, col_index):
+        """Check a value in a sample column."""
+        stripped_value = value.strip()
+        # do not check null values
+        # 'NA' is an allowed value. No further validations apply.
+        if stripped_value in self.NULL_VALUES:
+            return
+        # non-empty string
+        if len(stripped_value) == 0:
+            self.logger.error("Cell is empty. A categorical value is expected. Use 'NA' to indicate missing values.",
+                extra={'line_number': self.line_number,
+                'column_number': col_index + 1,
+                'cause': value})
+
+        return
+
+class GenericAssayBinaryValidator(GenericAssayWiseFileValidator):
+
+    """ Validator for files containing generic assay binary values.
+    """
+    def __init__(self, *args, **kwargs):
+        """Initialize the instance attributes of the data file validator."""
+        super(GenericAssayBinaryValidator, self).__init__(*args, **kwargs)
+
+    # (1) values defined in ALLOWED_VALUES
+    # (2) NA cell value is allowed; means value was not tested on a sample
+
+    ALLOWED_VALUES = ['yes', 'no', 'true', 'false'] + GenericAssayWiseFileValidator.NULL_VALUES
+
+    def checkValue(self, value, col_index):
+        """Check a value in a sample column."""
+        stripped_value = value.strip()
+        # do not check null values
+        # 'NA' is an allowed value. No further validations apply.
+        if stripped_value in self.NULL_VALUES:
+            return
+        if stripped_value not in self.ALLOWED_VALUES:
+            self.logger.error(
+                'Invalid generic assay binary value: possible values are [%s]',
+                ', '.join(self.ALLOWED_VALUES),
+                extra={'line_number': self.line_number,
+                        'column_number': col_index + 1,
+                        'cause': value})
+
+        return
+
 # ------------------------------------------------------------------------------
 # Functions
 

core/src/test/java/org/mskcc/cbio/portal/scripts/TestImportGenericAssayData.java

@@ -98,7 +98,7 @@ public void testImportGenericAssayData() throws Exception {
 
         // import data and test all mutational signatures were added
         ImportGenericAssayEntity.importData(file, GeneticAlterationType.GENERIC_ASSAY, "name,description", false);
-        assertEquals(61, getNumRecordsForGenericAssay());
+        assertEquals(60, getNumRecordsForGenericAssay());
 
         // test wether a record can be retrieved via stable id 
         GenericAssayMeta genericAssayMeta1 = DaoGenericAssay.getGenericAssayMetaByStableId("mean_1");

core/src/test/java/org/mskcc/cbio/portal/scripts/TestIntegrationTest.java

@@ -349,19 +349,19 @@ public void testLoadStudyEs0() throws Throwable {
             assertEquals("this is an optional custom case list", customSampleList.getName());
 
             // ===== check mutational signature
-            String testMutationalSignatureStableIds = "Nmut";
+            String testMutationalSignatureStableIds = "mean_1";
             String testMutationalSignatureMolecularProfileIds = "study_es_0_mutational_signature";
             assertNotNull(DaoGeneticEntity.getGeneticEntityByStableId(testMutationalSignatureStableIds));
             // ENTITY_STABLE_ID name description TCGA-A1-A0SB-01 TCGA-A1-A0SD-01
             // TCGA-A1-A0SE-01 TCGA-A1-A0SH-01 TCGA-A2-A04U-01 TCGA-B6-A0RS-01
             // TCGA-BH-A0HP-01 TCGA-BH-A18P-01
-            // Nmut ... ... ... 18	3	32	13	3	4	1	7
+            // mean_1 ... ... ... 0.370266873	0.010373016	0.005419294	0.022753384	0.037687823	0.016708976	0.100042446	0.104214723
             GenericAssayService genericAssayService = applicationContext.getBean(GenericAssayService.class);
             List<GenericAssayData> mutationalSignatureData = genericAssayService.fetchGenericAssayData(Arrays.asList(testMutationalSignatureMolecularProfileIds),
                     Arrays.asList("TCGA-A1-A0SB-01", "TCGA-A1-A0SH-01"), Arrays.asList(testMutationalSignatureStableIds), PersistenceConstants.SUMMARY_PROJECTION);
             assertEquals(2, mutationalSignatureData.size());
-            assertEquals("18", mutationalSignatureData.get(0).getValue());
-            assertEquals("13", mutationalSignatureData.get(1).getValue());
+            assertEquals("0.370266873", mutationalSignatureData.get(0).getValue());
+            assertEquals("0.022753384", mutationalSignatureData.get(1).getValue());
 
             // ===== check GSVA data
             // ...

core/src/test/resources/data_mutational_signature.txt

@@ -1,5 +1,4 @@
 ENTITY_STABLE_ID	name	description	TCGA-A1-A0SB-01	TCGA-A1-A0SD-01	TCGA-A1-A0SE-01	TCGA-A1-A0SH-01	TCGA-A2-A04U-01	TCGA-B6-A0RS-01	TCGA-BH-A0HP-01	TCGA-BH-A18P-01
-Nmut	Nmut	number of mutations	18	3	32	13	3	4	1	7
 mean_1	mean_1	mean_1	0.370266873	0.010373016	0.005419294	0.022753384	0.037687823	0.016708976	0.100042446	0.104214723
 mean_10	mean_10	mean_10	0.002709404	0.009212318	0.002650657	0.005475484	0.074175715	0.033049207	0.027323826	0.008861145
 mean_11	mean_11	mean_11	0.006035782	0.010095773	0.011926486	0.010637541	0.012168938	0.006641113	0.025730547	0.020463421

core/src/test/resources/data_mutational_signature_new.txt

@@ -1,5 +1,4 @@
 ENTITY_STABLE_ID	name	description	TCGA-A1-A0SB-01	TCGA-A1-A0SD-01	TCGA-A1-A0SE-01	TCGA-A1-A0SH-01	TCGA-A2-A04U-01	TCGA-B6-A0RS-01	TCGA-BH-A0HP-01	TCGA-BH-A18P-01
-Nmut	Nmut	number of mutations	18	3	32	13	3	4	1	7
 mean_1	mean_1	new mean_1	0.370266873	0.010373016	0.005419294	0.022753384	0.037687823	0.016708976	0.100042446	0.104214723
 mean_10	mean_10	mean_10	0.002709404	0.009212318	0.002650657	0.005475484	0.074175715	0.033049207	0.027323826	0.008861145
 mean_11	mean_11	mean_11	0.006035782	0.010095773	0.011926486	0.010637541	0.012168938	0.006641113	0.025730547	0.020463421

core/src/test/scripts/test_data/data_generic_assay_empty_cell.txt

@@ -0,0 +1,4 @@
+ENTITY_STABLE_ID	name	description	url	TCGA-A1-A0SB-01	TCGA-A1-A0SD-01	TCGA-A1-A0SE-01	TCGA-A1-A0SH-01	TCGA-A2-A04U-01	TCGA-B6-A0RS-01	TCGA-BH-A0HP-01	TCGA-BH-A18P-01
+mean_1	mean_1	mean_1	url_1	category_1	category_2	category_2	category_2	category_1	category_1	category_3	category_4
+mean_2	mean_2	mean_2	url_2		category_1	category_1	category_2	category_1	category_2	category_3	category_4
+mean_3	mean_3	mean_3	url_3	category_2	category_1	category_2	category_2	category_1	category_2	NA	category_3

core/src/test/scripts/test_data/data_generic_assay_valid_binary.txt

@@ -0,0 +1,4 @@
+ENTITY_STABLE_ID	name	description	url	TCGA-A1-A0SB-01	TCGA-A1-A0SD-01	TCGA-A1-A0SE-01	TCGA-A1-A0SH-01	TCGA-A2-A04U-01	TCGA-B6-A0RS-01	TCGA-BH-A0HP-01	TCGA-BH-A18P-01
+mean_1	mean_1	mean_1	url_1	true	false	false	false	true	true	yes	no
+mean_2	mean_2	mean_2	url_2	false	true	true	false	true	false	yes	no
+mean_3	mean_3	mean_3	url_3	false	true	false	false	true	false	NA	yes

core/src/test/scripts/test_data/data_generic_assay_valid_categorical.txt

@@ -0,0 +1,4 @@
+ENTITY_STABLE_ID	name	description	url	TCGA-A1-A0SB-01	TCGA-A1-A0SD-01	TCGA-A1-A0SE-01	TCGA-A1-A0SH-01	TCGA-A2-A04U-01	TCGA-B6-A0RS-01	TCGA-BH-A0HP-01	TCGA-BH-A18P-01
+mean_1	mean_1	mean_1	url_1	category_1	category_2	category_2	category_2	category_1	category_1	category_3	category_4
+mean_2	mean_2	mean_2	url_2	category_2	category_1	category_1	category_2	category_1	category_2	category_3	category_4
+mean_3	mean_3	mean_3	url_3	category_2	category_1	category_2	category_2	category_1	category_2	NA	category_3

core/src/test/scripts/test_data/data_generic_assay_valid_continuous.txt

@@ -0,0 +1,4 @@
+ENTITY_STABLE_ID	name	description	url	TCGA-A1-A0SB-01	TCGA-A1-A0SD-01	TCGA-A1-A0SE-01	TCGA-A1-A0SH-01	TCGA-A2-A04U-01	TCGA-B6-A0RS-01	TCGA-BH-A0HP-01	TCGA-BH-A18P-01
+mean_1	mean_1	mean_1	url_1	NA	0.010373016	0.005419294	0.022753384	0.037687823	0.016708976	0.100042446	0.104214723
+mean_2	mean_2	mean_2	url_2	0.002709404	0.009212318	0.002650657	<0.005475484	0.074175715	0.033049207	0.027323826	0.008861145
+mean_3	mean_3	mean_3	url_3	0.006035782	>0.010095773	0.011926486	0.010637541	0.012168938	0.006641113	NA	0.020463421

core/src/test/scripts/test_data/data_generic_assay_with_non_numerical_data.txt

@@ -0,0 +1,4 @@
+ENTITY_STABLE_ID	name	description	url	TCGA-A1-A0SB-01	TCGA-A1-A0SD-01	TCGA-A1-A0SE-01	TCGA-A1-A0SH-01	TCGA-A2-A04U-01	TCGA-B6-A0RS-01	TCGA-BH-A0HP-01	TCGA-BH-A18P-01
+mean_1	mean_1	mean_1	url_1	NA	0.010373016	0.005419294	0.022753384	0.037687823	0.016708976	0.100042446	0.104214723
+mean_2	mean_2	mean_2	url_2	NON_NUMERIC	0.009212318	0.002650657	0.005475484	0.074175715	0.033049207	0.027323826	0.008861145
+mean_3	mean_3	mean_3	url_3	0.006035782	0.010095773	0.011926486	0.010637541	0.012168938	0.006641113	0.025730547	0.020463421

core/src/test/scripts/test_data/data_generic_assay_with_not_defined_data.txt

@@ -0,0 +1,4 @@
+ENTITY_STABLE_ID	name	description	url	TCGA-A1-A0SB-01	TCGA-A1-A0SD-01	TCGA-A1-A0SE-01	TCGA-A1-A0SH-01	TCGA-A2-A04U-01	TCGA-B6-A0RS-01	TCGA-BH-A0HP-01	TCGA-BH-A18P-01
+mean_1	mean_1	mean_1	url_1	true	false	false	false	true	true	yes	no
+mean_2	mean_2	mean_2	url_2	false	NOT_DEFINED	true	false	true	false	yes	no
+mean_3	mean_3	mean_3	url_3	false	true	false	false	true	false	NA	yes

core/src/test/scripts/test_data/data_generic_assay_without_entity_id_column.txt

@@ -0,0 +1,4 @@
+name	description	url	TCGA-A1-A0SB-01	TCGA-A1-A0SD-01	TCGA-A1-A0SE-01	TCGA-A1-A0SH-01	TCGA-A2-A04U-01	TCGA-B6-A0RS-01	TCGA-BH-A0HP-01	TCGA-BH-A18P-01
+mean_1	mean_1	url_1	0.370266873	0.010373016	0.005419294	0.022753384	0.037687823	0.016708976	0.100042446	0.104214723
+mean_2	mean_2	url_2	0.002709404	0.009212318	0.002650657	0.005475484	0.074175715	0.033049207	0.027323826	0.008861145
+mean_3	mean_3	url_3	0.006035782	0.010095773	0.011926486	0.010637541	0.012168938	0.006641113	0.025730547	0.020463421