helmholtzai-consultants-munich / oligo-designer-toolsuite Goto Github PK

Adapt scrinshot pipeline to new toolsuite structure

• check if SCRINSHOT pipeline has all required property filters
• check if SCRINSHOT pipeline has correct blast specificity parameters, e.g. ligation region filter
• check if SCRINSHOT pipeline has correct blast corss-hybridization parameters
• check if SCRINSHOT pipeline has correct set scoring parameters
• split config into user and dev config
• define default parameters for each config files

• add transcript score:
  • transcript score = 1 - norm transcripts
  • number of transcripts that are targeted by the probe
  • normalize to 0-1 with 1=max transcripts per gene
• add score

Fix issue with exact match filter:

• Don't use eaxct macth filter for cross hybridization because we would need to compare to the reverse complement sequence, which cannot be done with the current implementation that makes the filter very fast. Instead one can simply use the Bowtie Filter.
• Only provide one apply function for the exact match filter, which takes a filter policy as input by which the duplicated oligos are removed, i.e. don't just remove all of them but have options on how to remove (e.g. by larger number of oligos)

The exact match filter already aggressively deleting oligos. The cross hybridization filter wouldn't remove anything if the exact match filter was already used

• test_sequence_generator_and_database:: test_oligo_database
• test_sequence_generator_and_database::test_ftp_loader_ensemble
• test_sequence_generator_and_database::test_region_generator_ncbi
• test_sequence_generator_and_database::test_region_generator_ensemble
• test_oligo_selection.py::test_oligosets_generation
• test_utils.py::test_data_parser FAILED
• test_utils.py::test_GFF_parser FAILED
• Need to add additionally: test_fasta_parser in test_utils

When we create mutile databases the filename for regions with insufficiemnt oligos will always be the same and mixed up for different databases → what happens when we define two database objects?
Solution: give it the name of the database variable → Create different names depending on database variable names

we start having many "calculate" functions in the database class which repeatedly add new attributes to the database. I would suggest to move the calculations of those attributes to a seperate class and then just call those classes in the "calculate" fucntions in the database. I would also suggest to create an attribute aclculator for each attruibute we add in the property filter class. In the property filters we can then call those attribute calculation classes which allows us to also compute attributes without running the filters. Suggestion of attributes to move into this class:

• GC content
• melting temp
• padlock arms
• isoform consensus
• number of transcripts
• oligo length
• seed region
• secondary structure
• homodimer length

Use effidict instead of the normal dictionary to store oligos

Currently the pipeline filters out genes that have fewer then config["min_probes_per_gene"] probes directly after finding possible probe candidates. However these probe candidates can overlap and the actual number of probes we can select could therefore be lower. We need to put that filter after calculating the overlap matrix.

Use Codecov to automatically collect code coverage reports

• Split types and constants file
• change types_seq to enums

Sometimes in the generated probes txt files I encounter multiple entries in the columns start and end, e.g. for TTLL10 (source=ncbi):

I've only seen multiple identical entries so far. Can there also be different ones? If not I assume it's not the intended behaviour, right?

When running get_overlap_matrix I also get an error probably because of this.