November 25, 2009 - Sequence Curation Flags
Back to Caltech documentation
In the future (thank you, Raymond):
starting the next conference call in December, please follow the new instructions:
8:30 AM Pacific Time
- EFFECTIVE DEC 2009
- tel. USA/Canada 866-528-2256
- tel. UK 0808-234-3475
- tel. (caller paid) 216-706-7052
- participant code 714646
Location: wherever you choose
Contact Person: Raymond Lee
Contact Email: firstname.lastname@example.org, 626-310-1144
Caltech - Karen, Kimberly, Paul, Ruihua
Sanger - Paul, Gary
WashU - John, Phil, Tamberlyn
Pipelines and options for flagging
Data may come into WormBase via various pipelines (e.g. Genbank, Knockout Consortia, user submissions) but for flagging data in published papers, here are the current pipelines:
- Curated papers are the best training set. Flagged papers can be used, if flagging was generally consistent.
- Curation status form has lists, but not completely up-to-date.
- How to evaluate the results of the SVMs
- Precision - Of the returned positives, how many are true positives?
- Recall - Of the true positives, how many were returned? (need to look at negatives for particular data type)
- What precision and recall values are acceptable?
- Is there any way for curators to see a list of features for each SVM? May help with understanding false positives.
- pattern matching
- category searches
- Curators need to tell Juancarlos they'd like to receive emails when authors flag a data type.
- Caltech needs to supply list of papers flagged since September 2009.
- Stats on return rates as of November 12, 2009 (supplied by Juancarlos):
Since Sept 1st, we have sent out 195 requests, and gotten back 72 results (36.9%).
Since Oct 1st, we have sent out 147 requests, and gotten back 52 results (35.3%).
Since Nov 1st, we have sent out 18 requests, and gotten back 7 results (38.9%).
|Flag name||Flag information||Number of papers flagged manually (from curation status form)||Flag email (from first pass form)||Getting author flags?||Approach||Curator(s)||Comments||Current pipeline sufficient?|
|gene symbol||newly cloned gene or new name for previously known gene||342||genenames, vanauken||no-vanauken||SVM (see comments)||Kimberly, Mary Ann?||Currently being combined with seqchange. Could possibly employ secondary screen with categories.||Still being assessed.|
|mapping data||genetic mapping data||194||genenames||possibly keywords or Textpresso categories|
|sequence features||regulatory sequence features, includes promoters, enhancer, elements in mRNA||248||worm-bug, stlouis, xiaodong (xdwang)||possibly keywords or Textpresso categories||SVM tried, but recall was low|
|mass spectrometry||mass spec analysis||65||gw3, worm-bug||keywords or Textpresso categories||Ruihua, Gary||will likely be a small number (10?) of papers each year|
|structure correction||gene structure corrections (see comments)||333||worm-ticket, worm-bug||re-try SVM with new training set provided by Paul Davis||Gary, Paul Davis||Ideally divided into four categories: a change in a gene's structure, the addition of an isoform, a change to one of the SL1/SL2 or polyA site features, a sequence correction in the N2 reference genome|
|sequence change||sequence of mutant alleles||981||genenames||SVM||Kimberly has assessed some, will put on wiki, anyone from Sanger interested in checking?||Being assessed.|
|new SNPs||new polymorphisms||50||tbieri||Most new SNP information comes via user submission emails. Can remove this data type from the pipeline.|
- Sanger and St. Louis should now be getting emails for any author flags for structcorr.
Below are the other addresses that got emails from manual first pass curation. If you'd like to get emails when authors flag this data type, please confirm by putting Yes or No next to the email name.
- genesymbol - genenames
- mapping data - genenames
- sequence features - worm-bug, stlouis
- mass spec - gw3, worm-bug
- seqchange - genenames
Author-flagged papers since September
- Get list of papers flagged for sequence-related data by authors since September from Juancarlos.
- genesymbol (10 papers)
- mapping data (2 papers)
- sequence features (2 papers)
- mass spec (1 paper)
- structcorr (4 papers)
- seqchange (11 papers)
- new SNP - no author flags
Back to Caltech documentation