WormBase-Caltech Weekly Calls January 2018
- 1 January 4, 2018
- 2 January 11, 2018
- 3 January 18, 2018
- 4 January 25, 2018
January 4, 2018
- Citace upload to Wen, Tuesday January 16th, by 10am PST
- Upload to Hinxton on Jan 19th
Strain data import to AGR for disease
- Will begin to consider pulling in strains into AGR
- Will need to think about how genotypes are built and stored at other MODs
- We should encourage authors to include strain IDs
- Diseases are annotated to genes, alleles, and strains within WB
Curating phenotypes and diseases to strains or genotypes
- Should we generate a ?Genotype class to capture genotypes without a known strain name? Or to capture relevant/relative genotypes thought to be responsible for a phenotype or disease?
- We could create un-named strain objects, that use a new unique identifier as a primary identifier and represent the entire genotype of a strain used
- Introduction of a new ?Strain class attribute of a unique serial identifier (like WBStrain00001) would be very costly to implement; would need to consider how crucial this is before implementing
- We can, instead, use new strain (public) names like "WBPaper00012345_Strain1", etc. instead of creating new unique ID attribute for un-named strains
- When curating phenotypes to strains, we will want to specify what is the relevant/relative genotype that is causative/correlated with the disease or phenotype observation
- Would be best if the specification of the relevant genotype used controlled vocabularies (when possible) and free text (when needed); would need to work out the logistics/mechanics of such curation
- Transgene-phenotype curation currently specifies causative gene, but would be more complicated for strains
- Alternatively, we could create the ?Genotype class to represent the abstract "relative"/"relevant" genotype thought to be responsible for the phenotype or disease, and annotate directly to that ?Genotype object
- ?Strain approach:
- Use strain if named (but important to know if the control strain is not simply N2)
- If control strain is simply N2, causative genotype (and respective components) can be inferred from strain genotype
- If control strain is not N2, causative genotype and components would need to be specified at the moment of phenotype/disease curation (by mechanism to be worked out)
- If no strain name provided, create "un-named" strain that contains the entire genotype provided by authors
- Control strain issues above would still need to be addressed
- Use strain if named (but important to know if the control strain is not simply N2)
- ?Genotype approach:
- ?Genotype class could represent individual instances of relevant/relative genotypes that are suggested to be causative for a disease or phenotype
- ?Genotype objects would be created with formal construction, with DB associations to each component object (e.g. alleles, transgenes, etc.) as well as free text descriptions (for components with no corresponding DB object)
- Such ?Genotype objects could be used repeatedly throughout a paper when applicable, but would likely not be used in any other papers (we would likely accumulate redundant objects in the DB)
- We may want to consider strains with same public name that have diverged
- Apply new strain names with prefixes/suffixes? Create new strain objects? Keep original?
- Need to determine how each AGR member DB curates phenotypes or diseases to genotypes: is each "genotype" a relative or absolute genotype?
January 11, 2018
- Eppendorf tube openers with WormBase logo?
Update on AFP Form
- Idea is to move from author flagging to author validation of text mining and data submission wherever possible
- Goal is to flag all data types in a paper and either curate at WB or share with a group that does curate that data
- SVM flags and author flags can/will be used as filters in TPC
- Provide examples of what we want for each type of data to help avoid confusion
- Recognize entities automatically and show list to author
- Species, strains, genes, alleles, transgenes, etc.
- Ask to verify or add unrecognized
- Could show known/existing objects with checkboxes
- Possibly include unrecognized pattern matching objects? Ask author to verify if these are real?
- For strains:
- Show recorded genotype for verification; maybe ask to update/modify if needed?
- For transgenes:
- When author submits new transgene, send them to a transgene form, or send them an email asking for details?
- Form could be for both strain and transgene
- Mapping data: still ask for? Maybe for balancers, but no one is reporting that. Could still ask if there's interest
- Maybe provide option for author to save their progress and return to the form later
- Ask for allele, RNAi and overexpression phenotypes with links to Phenotype form
- Also ask for drug/chemical and environmental perturbations (call treatment?); store as free text for now, accommodate with new data model when available
- Gene site- and time-of-action, mosaic
- Appears to be confusion from authors about mosaics. Should we keep this?
- Will keep gene site-of-action and time-of-action; leave unchecked (no SVM, yet) but allow users to indicate
- Cell and anatomy data
- Cell function ("Cell ablation (laser/genetic) data, optogenetics")
- Ultrastructural analysis
- Interaction data
- Genetic interactions
- Physical interactions
- Functional complementation
- Comparative genomics
- Gene expression & regulation
January 18, 2018
- May be good to get (possibly anonymous) written questions or suggestions after presenting
- Wen will have Skype call with Yishi Jin
- how do we peer-review single experiment? No supporting information to corroborate a larger story
- Is the greater benefit of peer-review that the whole story is assessed by reviewers
- Do MPs help or hurt reproducibility?
- Larger papers may have lots of poor experiments that don't get much attention but still pass peer review
- Dedicated peer review on single experiment may be more rigorous
- What are the criteria/minimal requirements to micropublish?
- Concise descriptions
- SimpleMine has multiple descriptions output; people asked about the different types
- Yishi Jin suggested that we remind users to update manually written descriptions
- Showing last-updated date is important
- Automated descriptions relies on primary data; will rely on forms and community submissions
- Microreviews? Would want to guide authors what data we want; provide a template?
- Public/community education issues
- Users shouldn't assume that WormBase is comprehensively up to date
- Wen will also present at MidWest meeting (Ann Arbor, MI) in April and Boulder, Colorado in May
- Will assess topic interest ahead of time
New Cytoscape display for interactions
- Sibyl developed a new Cytoscape display for interactions, now live with WS262 release
- Simplified colors and subtypes
- Redraw button to clean up the graph based on what you want to see
- Play around and let Sibyl and/or Chris know about issues
January 25, 2018
UCSF visit report
Questions from the audience
- Is there a way in WB to pull out verified CRISPR guides?
- No, WB does not currently curate these
- Issues: some are more general, most are very specific
- Don Moerman has site
- Single cell RNAseq from Waterson lab in WB? Paper is in WB, person-paper connections might be on staging, Kimberly verifying this. Gary W will put data in WS264. Comments from Gary on this paper:
‘The main problem is that the authors can sort the cells into groups (corresponding to tissues, I think) and sometimes sort them into single cells, but it is hard to identify the tissues or cells. I think they found 29 groups, of which about 20 appear to be single cells. I think they have a website where they invite other researchers to make suggestions about how to identify cells in their data. Currently I regard this data set as still undergoing analysis and I'm waiting to see if they improve the deconvolution and identification of the cells. The RNASeq data from this paper will be going into the current Build (WS264) but I will not be adding it to SPELL this Build because displaying it probably requires more thought. I'm not sure that tools for displaying single-cell data have been developed very much yet. There is potentially a lot of information if eventually all 959 or 1031 somatic cells are displayed!
- Not entirely complete, but very useful
- How do we identify which cells to annotate the data to, reliably? Take 1st order approximation for now; work to improve going forward
- List of fem-3 alleles excluding natural variants in WormMine, how to do that? The template should be fixed (checkbox)
- WormMine template was broken; model has been rolled back to WS260; working now. Here's the query:
<query name="" model="genomic" view="Gene.primaryIdentifier Gene.secondaryIdentifier Gene.symbol Gene.alleles.primaryIdentifier Gene.alleles.symbol" longDescription="" sortOrder="Gene.secondaryIdentifier asc" constraintLogic="B and A"> <constraint path="Gene.symbol" code="B" op="=" value="fem-3"/> <constraint path="Gene.alleles.naturalVariant" code="A" op="=" value="false"/> </query>
- Is the increase of published paper due to an increased number of labs? Expansion of the field? Is there a correlation between the increased number of publications and increased number of labs
- One user found powerful to be able to use the Galaxy server to do analysis after exporting data from WormMine
- Can you do protein domain analysis with WormMine? Is the protein->motif precanned query the best option?
- Enrichment: how to see which genes are expressed in a tissue or cell -> pointed user to the ontology browser
Skype call with Yishi Jin and Sreekanth Chalasani
Participants: Daniela, Karen, Wen, Jin, Shrek
- Not all images available due to publisher agreement -> not clear to users, we should put a disclaimer somewhere
- Shrek: thinks gene expression display should improve => hard to figure out all expression patterns
- Shrek and Jin feel the most important info is the reagent, and that should be displayed on the gene page/expression widget other than the Expression page
- Images are identical -> example on the eat-4 expression page. This is normal since one image depicts multiple neurons and the association Anatomy neuron applies.
- On the eat-4 expression page the problem is amplified as one picture shows 70+ neurons
- We should remove the image column and have links to images only on the panel above (as currently displayed)
- The image column should be replaced with reagent and description, if possible. Will need to talk to Sibyl and see what is duable.
- Shrek pointed out that the eat-4 concise description is out of date, we explained that in the near future the manual descriptions will be superseded by automatic descriptions
- Neuron connectivity: missing neuron connectivity pages, there is a new reconstruction of neuroanatomy from david hall, would be great to integrate
New Framework for paper classification
Valerio will present a new project aimed at creating new paper classifiers: https://github.com/valearna/tpclassifier
- A new classification pipeline will be set up and run in parallel to the current SVM pipeline to compare the results
- Results of new classifiers trained with the original training set of the current SVMs and tested on papers validated on the curation status form: https://docs.google.com/document/d/1MMJ3J2H6he_T1cKQK5wDA0USXbJqez6EWhmJqEFsNjE/edit?usp=sharing
Proposed graphic interface for 5'UTR of ORFs
April will present an idea for a graphical interface for the 5'UTR of ORFs, that can be used to display gene regulation and interaction (eg. transcription-factor binding sites) for genes.
see this link for a pdf describing the idea: https://drive.google.com/open?id=0B1vRTZbiAp0pZndLUDdSU3BWTUt1OTIzWTc3OXBwRHh3QXQ4
- Paul S. attended workshop
- Funding source changes
- Interest in genetic/drug interactions
- How to assess curation coverage?
- Interest in automating MESH term association to papers
- Could use SVMs, turn-around time may be limiting factor (maybe not anymore)