2015-09-08 Tuesday Team Meeting
Dundee: Will, Petr, Kenny, Dominik, Graeme, Mark, Balaji, Helen, Remote: Josh, Sebastien, Balint, Gus, Jean-Marie, Ian, Josh B, Colin, Wilma, David, Chris, Melissa, Stick, Kelli
Agenda - 2:30pm Start
Accepting minutes from last meeting
Project Timelines (2-3 minutes each)
OMERO
- 5.1.4 (Seb) Couple of stories progressing: wsgi mostly done, used in daily ci builds CLI bug-fixes mostly done ImageJ fixes and Group support for matlab 90 tickets open in trac webstart deprecation announcement in preparation. JM: open PR today Jean-Marie: where to push tickets out of 5.1.4? Seb: 5.2.0 is fairly well defined. Push to 5.x or unscheduled.
- 5.2.0 (J-m) JM: Cleaning of jdoc - prepare for major changes. Pojo migration Petr: looked good last week. Last bugs from testing have been fixed JM: Balaji is extending robot tests, based on jstree branch
- Regions v2 (Josh) Josh: try to get a minor change in (XML) - new repository. Petr: ICY guys interested in ROI collaboration. Josh: added to trello
Bio-Formats
Glencoe Update (Chris) Discussing ROIs
Repositories
- Virtual Microscope/teaching Gus: Presentation in Glasgow went well. VM easy to use compared to their systems, but lacks e-learning tools. A couple of people interested - they will get in touch.
- IDR Project Josh: data getting imported - Dozens of plates, Many TB. Eleanor: getting used to new demo system. Loading data.
FLIMfit/OPT (Ian)
- No response as yet to my post about building the ThunderSTORM imagej plugin. Does someone with better maven skills want to try in case I'm missing something?
- Can connect to OMERO with Matlab-based localisation tool. Can read OMERO.table and create an image.
Other consortium timelines (5 minutes) Petr: Madison trip went well. 30-40 people in workshop. Mostly quite impressed, but quite stressful. No list of participants. Users interested in saving IJ workflows.
Presentation by Balint Antal on Mineotaur
- Interactive visual analytics for high-content microscopy screens
- web based, visual analysis for biologists
- sharing of research findings, access and analysis of ‘big data’
- Architecture: Java-based web server & graph database queried by rest API from js in browser.
- Graph database better than RDBMS for analysis. Relational joins a problem (slow)
- Use Neo4J. Each E.g. gene is a data point. Connected by edge.
- E.g. 7.3 million nodes, 131 features for each cell
- Terabytes of image data, 100 GB of raw feature data (30 GB after filtering)
- Model: Microtubule, Cell, Strain, Gene, Phenotype nodes etc.
- Server: spring framework (Java), Neo4j, thymeleaf template engine.
- Usage, select feature and aggregation E.g. length, average. Then filter by E.g. cell cycle stage, strain etc.
- Backend starts at gene name, filters phenotypes by cell cycle stage etc.
- UI design: Aim is to provide access to all info with small number of clicks, easy to learn
- Once the data is queried, much analysis can be done in the browser:
- Inspect data points, plot subset, filter by hit type, export graphs etc.
- Client side: Use requireJS to provide modularised architecture.
- Import data from text files: Feature data csv, Labels data 1/0 matrix.
- Java classes generated on the fly -> proper validation of the input
- Identify duplicates from complex rules. Classes not reused.
- OMERO + Mineotaur: Load features and labels from Screen attachments.
- Render images from OMERO (previously just linked to OMERO web app)
- Next: Publish a paper
- Extend model to spatio-temporal data
- Eleanor - are hit-type lined to cells or Genes? Balint: Genes. Eleanor: Not sure we have that info for our IDR datasets.