WormBase-Caltech Weekly Calls March 2011
March 3, 2011
Every other month patch:
- We will try to generate an ACEDB patch every other month, starting with the Paper class data
- Need to coordinate with Juancarlos and Todd
Need to check .ACE patch files before upload:
- Will need to be vigilant about checking for errors and inconsistencies before sending to Todd to put up on website
- Curators need to check their own data for problems
- Wen will check for consistency between the different data types
Think about connecting website to Postgres:
- Think about showing Postgres data "immediately" on site
Delayed release cycle:
- Will require more work to prepare for more frequent release of certain data types
- Aside from Kimberly's data, most data types are not urgent (e.g. Expression pattern)
- What are the users feeling?
- Having data faster will help users; they don't ask, because they don't see it
- On-the-fly updating of website? Like Postgres?
- Since we use ACEDB, we have to patch WS with .ACE file, or rebuild whole thing
- Flat file Postgres database, replaced every night?
- Website calls Postgres directly for certain data types?
- Performing build without sequence is easy? Do everything without sequence?
- How to integrate sequence data with other data once they're decoupled through the patching process?
- We need .ACE patch files
- Concise description separate from most else (but connected to papers)
- Do papers first?
- Website can show anything
- If we have a lot of patches, will not have check for data inconsistency/confliction
- Trial patch .ace files for papers first
- Juancarlos: Scripts that check differences between data dumps; scripts are data type specific
- Curators need to talk to Juancarlos about the importance of different data tags
- Paper .ACE file: Would include bibliographic info, journals, authors, genes associated from abstract or added manually
- One reason for more frequent releases: because we have first pass author forms; show them we add it quickly
- what will be added through the forms: expression patterns? RNAi (difficult?)?
- We should check patch before we send to Todd!!! Don't want to crash database
- How frequently to patch? Weekly? Daily? Check with Todd, how often he can load them?
- Chron job to create patch ACE files, send to curators to check for problems, then send to Todd
- Interdependency of data types; curators rely on other curators?
- Postgres directly to website? Todd would have to work it out
- New information flag on website? Toggle visibility?
- How do we know that the data do not conflict with each other?
- What are common problems? Dumper script goes bad, makes broken lines, empty fields
- Error catching mechanisms? More checks on postgres? Dump files?
- Data merging problems? What are the cases that are conflicts? Prevent them? Know beforehand?
- If we don't know, as long as it doesn't crash the database or fail to load, then OK
- Don't do -D stuff, maybe? No deletions? Skip typos?
- Always have to check ACE files anyway, but have to do every week (2 weeks?)
- We can try a patch every other month
- What can we do without the patch?
- Did SAB talk about changing to relational databases?
- Get website going as is first, and see if it matters?
- If people don't want to change data models, we can switch over to relational
- Separate panel on website directly from Postgres?
- Wen can check the data integration every other month for patch
March 10th, 2011
Release schedule and patches
- What is the appropriate frequency?
- Scheme: do what we're already doing, Wen merges into citace
- Excluding sequence related data
- Need to include Mary Ann's data (strains etc.)
- Daily update too frequent; maybe once per month/week
- Submit .ACE file to Todd with simple syntax; easily parsable; old description removed and new information added
- Make updates only in contrast to last WS, not previous patch/temporary upload
- ACEDB diff step only relative to WS
- Wen: Postgres can dump diff ace file; already have diff ace files for every data type at Caltech; integrate into citace;
- Raymond: integration is important; we need to talk about how much work needs to be done by each approach
- Wen: consistency checks, backups, store each version?
- Rayomnd: citace 224 to 225 (for example), display done on class level
- Example: Gene page; only update information relevant to gene class to be displayed
- Do once per month: faster than currently because it doesn't have to go through the dev site
- If we update Citace to Citace, missing a lot of cross-references?
- Mock citace with Mary Ann's data? Becomes diff base; Mary Ann submits (non-sequence related) data directly to WBCIT
- Build low-connectivity ace at WBCIT? Add Mary Ann's data, remove RNAi
- Todd: important consideration: things added won't be available for search until formal release; weird things about diff; new reference associations with genes; a lot of duplications?
- Raymond: will look at it
- Todd: WBGene00000846, example, see how fast it loads, go from there; would like a single ACE file (concatenation of all individual ace files); would not happen on development; would have to happen on production releases; would take production database off line, clone it, and upload it to all production nodes
- Individual curators need to check their individual ace files for errors
- Frequency: monthly
- Todd: we should just run some tests first, to check feasibility
- Raymond: WBGene1, example, concise description has typo, WS225 has typo fixed from WS224, diff file shows:
- -D old_description
- Load diff ace into original database?
- parallel display; unrelated to resident WS?
- Todd: producing two web pages for each object?
- Raymond: No, only changing relevant tags, etc.
- Todd: Two databases running at same time inefficient; include timestamps?
- Raymond: No, cannot include timestamps
- Wen: Send patch ace files to Todd
- Raymond: In conflict with versioning; how to show new data
- Wen: Call it "WS225.1"
Human Disease Relevance tag in Concise Description
- Ranjana: Sent out e-mail; human disease tag "Human Disease Relevance"; to clean up concise description form (old tags in form outdated); could be putting more information into concise description; OMIM human disease
- Raymond: make not just text field, but make entity field pointing to object; meant to be human readable, this may break up the concise description into OMIM-related and OMIM-non-related info; why parse the data into a tag?
- Paul S: OMIM descriptions as a separate tag
- Raymond: Rewrite concise description?
- Ranjana: No
- Paul S: Would you mention human disease relevance in concise description? yes, but if it's just a link out to OMIM, then separate out; OMIM may have changed since Erich wrote original script; check OMIM for new information and tags that may be able to get pulled out
- Ranjana: Michael Paulini can consolidate orthology information?
Karen: Transgene model
- A lot of changes to propose
- Deletions of tags; more coming
- Other things in database connected to transgenes
- Many things in transgene objects that may be able to be parsed into different tags (new job for someone?)
- Strict nomenclature for transgene descriptor
- Clones present need to be parsed into clone class?
- Todd made Clone page;
- Start with vectors/backbones and then work on specific plasmids
Gene class-phenotype connections and descriptions?
GSA markup at Flybase
- Flybase is not willing to fully QC all papers
- Do we push Flybase and/or provide a better tool to QC?
- Are we worried about the GSA markup for flies not looking professional?
- People need to be willing to pay for the QC/curation; depends on database priorities
- CIT will spend more time on in-house development to make Fly GSA markup easier/more efficient
Putting SPELL on Amazon cloud?
March 17th, 2011
Transgene Information was the only topic discussed.
Karen noticed that CGC strain information contains more detailed information about transgenes then what is found in the papers that describe the transgenes. She would like to integrate the CGC strain information into the transgene curation pipeline. This could potentially lead to some redundancy, but the consensus was that this was a good idea. Going forward - Initially Karen will go through all the CGC strains that have transgene info and add it to the transgene objects in WB. Once that is accomplished we will institute some system (like the NBP allele info) to update as new strains w/transgene info are entered into the CGC.
March 24, 2011
Ranjana- Upload this week? Wen: No, next month
- Send info to Mary Alvarez
- register for housing after April 1st
Raymond: update on patch file
- Generate patch file at regular interval between 2 month formal releases
- Use patch file to update the web page
- Proposal for citace, use ACE diff utility -> citace intermediate
- Start with one upload between formal releases
- Submit patch file to OICR
- Testing patch file; submit to Todd for testing
- How to generate a good patch file?
- tried on one class; some issues
- patch file looks OK, sent to Kimberly
- test by reading into ACeDB
- -D bug; problem for hashed tags
- solution: separate -D lines from those without -D; read in to ACEDB as different blocks
- maybe 2 weeks to get working
- Gary: data linked to ontology terms?
- patching OBO files?
- Raymond: we assume that not all info is going to be updated mid-build
- have to accept some dead links
- Juancarlos: How does ACE diff handle ordered vs non-ordered tags?
- Raymond: Not sure; we'll have to test
- is this only an Author tag issue?
- Juancarlos: think there may be others; not sure
- Raymond: Alternative solution to ACE diff, acedb patch file from postgres directly
- Juancarlos: Possible to dump two consecutive ace files from postgres, write a script to compare them
- Raymond: what about data not in postgres? Anatomy? May be possible
- Raymond & Juancarlos: ACE diff approach is preferred
- Will ACE diff handle model changes? No
- Juancarlos: have to change parser everytime there's a model change
- Kimberly: will go through the diff file carefully
- Raymond: What Wen says counts; ignore robotic e-mails
Ranjana: Question about OA
- Dead genes in postgres show up in OA
- Gene table vs dead table; gene table contains everything
- Term info should indicate that a gene is obsolete
- In future, if something is wrong, it will not show
March 31, 2011
Next acedb upload - May 26th
Transgene model change proposals (Karen)
- Handouts (old vs new model)
- new model much shorter
- "Reporter_product" and "Integrated_by" now ?Text (rather than list of options)
- Change "Integrated_by" to "Integration_method"
- OA will provide drop down menu for controlled vocabulary of terms (e.g. "GFP")
- switch "Reporter" with "Reporter_product"
- "Isolation" now called "Construction"
- "Coinjection_marker" change from ?Transgene to ?Text
- Remove "Coinjected_with" and just keep "Coinjection_marker"
- Remove "Fragment Text" tag?
- "Location" now sub-tag of Construction, populated by Juancarlos' script
- Change "Location" to "Laboratory"? Todd: This would be helpful for web team
- Look at "Author" tag; obsolete? Use Laboratory (under location) instead?
- New tag "Genetic_information"
- Change "Transgene_type" to "Reporter_type"
- "Multi_point" should be indented
- Rescue ?Variation only; remove Rescue ?Gene
- Remove Coinjection... stuff from Used_for tag
- Add Associated_with tag
- Supporting data removed
- Where is "Is" vs "Ex" info stored? Currently in "Isolated_by" subtag "Not_integrated"
- Remove "Not_integrated"
- Should have more explicit display/capturing of extra-chromosomal vs integrated transgene
- Where do we store integration location information?
- Add to Genetic_information as "Extrachromosomal" and "Integrated" sub-tags
- "Driven_by" questions?
- "Driven_by_gene": should this be just primary driving sequence? Much more informative
- Yes, but very rarely is sequence available from authors
- Some of this info held in "Remarks" field
- Good if we could get primers, at least
- We will keep curating this data in the "Remark" field; maybe hire a summer student/work-study to perform mapping/sequence extraction
- Out-sourcing/crowd-sourcing curation?
- "Driven_by_gene": should this be just primary driving sequence? Much more informative
- Hopefully in database by WS226
- Clean Remark box for WS226 (not clean for WS225)
Citace Patch (Raymond)
- Patch file e-mailed out
- step forward for figuring out protocol for patch
- If supplemented with geneace, should cover everything outside of the sequence/genome browser domain
- How long will it take to load?
- Doesn't take long to load into citace
- Will test loading into full ACEDB
- File only a few MBs
- Todd: lots of XREFs slow things down
- Complete test will require Geneace patch from Mary Ann
Virtual Worm Site and You Tube tutorials (Chris)
SVMs for Genetics papers
- Genetics papers were missed for some reason
- Need to rerun Genetics papers through SVM
- Need to get through Textpresso as well (before SVM?)
- Need to clarify if those papers are on Textpresso-Dev
- Who is handling/should handle KOG data in WormBase?
- Anyone at Caltech have thoughts?
- Should get Erich's opinion
- Can we maintain it ourselves? Is that out of the question?
Updated changes for OA (Juancarlos)
Daniela and Juancarlos have updated old picture data
- Will send a couple of links around for people to review