2021-01-18 4DN-OME

--- tags: 4DN --- # 2021-01-18 4DN-OME Attending: Jason, Josh, Jean-marie, Xiaopeng, Yang, Jian Ma, Will Moore, Sarah Aufmkolk * Jason: call with center was illuminating. Center is triying to decide what it's going to do first, what cell lines, what data to collect. - Jian: Agreed. Just getting started. - Jason: difficult questions but can't be answered in an hour long call. See upcoming PI call. - Jian: data generation need is there. Dave Gilbert wants to make sure people have something to do. For table 3 there will be certain data types that will be generated regardless. But unclear which cell line to prioritize. Several factors involved: availability of public data. Biological question of interest to center. Feasibility of doing imaging & follow-up (re-wiring) experiments. - Jason: *but* specific decisions (cell line/data) need to be taken and until then we're in a holding pattern for visualization. Then we are in a position to discuss with Nicolai and Ting on which datasets. - Jian: can happen in parallel. (Genome) Imaging plan: Two types- superres in targeted region. Or lowres whole genome. Currently looking to do more targeted regions. Developing a new approach for whole genome. Couple of emerging methods. One lab could develop the library another could do the imaging. - Jason: have heard "I have data but I don't see a reason to share it since it's not useful to anyone". i.e. looking for data. - Jian: the original imaging data? Yes, with the locis. Jian: will discuss again with Long Cai. - Jason: on the imaging call last week saying, "we need to see the data" (led by Caterina). Only interesting when someone is prepared to transfer data somewhere for re-analysis. - Jian: Tom Misteli's published dataset perhaps. Will need to reach out to the labs. Sarah? - Sarah: was going to generate test but then there are errors. Have been trying for 3 weeks to get the data shareable. - Jason: people don't care about "reproducibility"; only making new science/products. What would be valuable for "us"? Key is that there's imaging data & loci. - Sarah: need to have a start. - Josh: some feedback (twitter/users) as soon as we have some good dataset. - Sarah: "one good dataset" is hard. - J-m: same as code, "not clean enough" - Jason: something we can start an experiment with. - Sarah: our data should definitely go out. - Jason: can put it in IDR as a "reference dataset" for going first. Can build jupyter notebook for that data. - Sarah: will definitely push on ours. Happy to go first. Localizations. (TB of original images) - Jian: want to prove to people that it will be useful. And then plan for which types of toolboxes will be valuable. - with Frank, have integrated analysis group meeting every 2 weeks on Thursdays. Good discussions - question of how to integrate OMICS & Imaging - Sarah: still questions of which data (e.g. whole genome might be tough) - Jian: HTP-FISH can also do chromatin and nuclear body interactions. Can go on show-case list. (Jason: had also talked to them) - Jian: what is the goal of the imaging group - Jason: trying to make it about data **exchange**, get data moving. People are using different methods for identifying the localizations, and there's no clear consensus. - Sarah: have often tried multiple algorithms, but there isn't always the time. - Jian: discussed with Peter, Sarah? - Sarah: don't know. Ting talks to a lot. - Jason: David, Caterina, Ting, Peter were all there discussing ("not sure what we're doing with imaging data.") - Jian: that was phase 1 - J-m: what we can immediately? Zarrify? Enables analysis. - Josh: definitely. May need to try several solutions. (Using files in the cloud to improve scaling) - Sarah: our data is very manual. Anything will get our blessing. e.g. assigning to clusters. All the data will be sequential. If you look at MERFISH, you'll have multiplexing and it will be troubling to assign everything. Need to encourage capturing the steps that need to be taken to assign images to loci. - JIan: Most promising might be Misteli. - Xiaopeng: Figure out how to apply genomics coordinates on the image. What's the scientific question we ask? Can already direct browser at the location. But that's it. - Jian: not just ensemble average distance but variablitity and dynamics. Ask how to represent that, and from many different images.... for 1000 cells. May be interest into looking into the various clusters. Can come up with a few of those. - X. need detail plan to make user experience feasible for this type of investigation. Trying to figure that out. - X. discussed with Josh that we could provide the data via microservice or distribute as a file in the cloud. Compose components into a pipeline. - Rough notes on Whiteboarding together (see screenshot): - https://en.wikipedia.org/wiki/Track_hub - Issues of update. - Multiscale information. - Promise function is a javascript return from the query service. * Summary (Jian) - Next step: go back to existing datasets and try to get access to them. - Keep technical conversation going & merge strategies to work with the datasets. - Jason: lot of potential interesting ideas to try. :+1: and submit data to IDR!