Add an option for ignoring individual samples when download a GEO series or SRA project

Author: arteymix

rnaseq_pipeline/sources/geo.py

@@ -136,8 +136,11 @@ class DownloadGeoSeries(DynamicTaskWithOutputMixin, DynamicWrapperTask):
     Download all GEO Samples related to a GEO Series.
     """
 
+    ignored_samples = luigi.ListParameter(default=[], description='Ignored GSM identifiers')
+
     def run(self):
         gsms = collect_geo_samples(self.input().path)
+        gsms = [gsm for gsm in gsms if gsm not in self.ignored_samples]
         if not gsms:
             raise ValueError('{} has no related GEO samples with RNA-Seq data.'.format(self.gse))
         yield [DownloadGeoSample(gsm, metadata=self.metadata) for gsm in gsms]

rnaseq_pipeline/sources/sra.py

@@ -182,9 +182,10 @@ def output(self):
 
 @requires(DownloadSraProjectRunInfo)
 class DownloadSraProject(DynamicTaskWithOutputMixin, DynamicWrapperTask):
+    ignored_samples = luigi.ListParameter(default=[], description='Ignored SRX identifiers')
     def run(self):
         df = read_runinfo(self.input().path)
-        yield [DownloadSraExperiment(experiment, metadata=self.metadata) for experiment, runs in df.groupby('Experiment')]
+        yield [DownloadSraExperiment(experiment, metadata=self.metadata) for experiment, runs in df.groupby('Experiment') if experiment not in self.ignored_samples]
 
 @requires(DownloadSraProjectRunInfo, DownloadSraProject)
 class ExtractSraProjectBatchInfo(luigi.Task):

rnaseq_pipeline/tasks.py

@@ -396,13 +396,15 @@ class SubmitExperimentBatchInfoToGemma(RerunnableTaskMixin, GemmaCliTask):
 
     resources = {'submit_batch_info_jobs': 1}
 
+    ignored_samples = luigi.ListParameter(default=[])
+
     def requires(self):
         # TODO: Have a generic strategy for extracting batch info that would
         # work for all sources
         if self.external_database == 'GEO':
-            return ExtractGeoSeriesBatchInfo(self.accession, metadata=dict(experiment_id=self.experiment_id))
+            return ExtractGeoSeriesBatchInfo(self.accession, metadata=dict(experiment_id=self.experiment_id), ignored_samples=self.ignored_samples)
         elif self.external_database == 'SRA':
-            return ExtractSraProjectBatchInfo(self.accession, metadata=dict(experiment_id=self.experiment_id))
+            return ExtractSraProjectBatchInfo(self.accession, metadata=dict(experiment_id=self.experiment_id), ignored_samples=self.ignored_samples)
         else:
             raise NotImplementedError('Extracting batch information from {} is not supported.'.format(self.external_database))