Derive a sample dataset

[ns-rse]

We require a sample dataset for this tutorial so that there is a standard set of images for users to process as they work through the different stages.

I realised whilst working on #11 that we probably don't want to copy over the files and directory structure used in the Unix Shell Software Carpentry and would be better served to have a set of AFM images in a hierarchical structure which we can use to demonstrate directory structure and navigation in the Command Line sections and how TopoStats scans for images across the base_dir and how that in turn influences the structure of output.

For this I feel we need more than just minicircles.spm that is used in the TopoStats test suite.

It would also be worthwhile including an image that doesn't process so that we can demonstrate what happens when this arises and how to handle such cases.

If anyone has images that are not under Non-Disclosure Agreements and is able to share them please provide links in reply to this issue (or email me the Google Drive link, no need to send the image).

[derollins]

Images from the NDP52 project are now all published into a repository so can probably be used for this purpose.

e.g.

SPM file: https://drive.google.com/file/d/15pfSigAkha5VWlaIVDgSA4pZ_F8ETm49/view?usp=drive_link

more files can be found here:
https://drive.google.com/drive/folders/1oVpwfpvVGQNnL6XBGl45hEjA0dGAVBV9?usp=drive_link

[derollins]

Additionally there is 'dirty' blanks that all experimentalists will have a good collection of and could be used for this purpose (they are useless otherwise!)

UPDATE (using edit ;) ): here is a folder of "interesting" blanks: https://drive.google.com/drive/folders/1-4rucSQqm1aY40rxd63e8nGEuAcHnRiA?usp=drive_link

[ns-rse]

Brilliant, thanks @derollins I'll have a look through next week when I'm back on training development.

Useless GitHub Tip : You can edit posts (pen icon at the top right of the gray bar above posts.

[ns-rse]

Another source of material is those that are available from the Minicircle work, see figshare.

This would be useful is it has both "super-coiled" (-3) and "relaxed" molecules imaged which would help show differences in the distribution of statistics that are calculated with each.

[MaxGamill-Sheffield]

Looking at minicircle ∆Lk 0 & -3 for a DNA example. Trying to replicate the following results (primarily looking at aspect ratio to show greater compaction as the DNA is under twisted).

The config file for this shouldn't be too fiddly as the sample is "nice".

Looking at NDP52 for a protein example. Trying to replicate the following results (primarily looking at the ferets, maybe volume too to visualise difference between monomers and dimers). "Peaks in KDE plots were used to determine particle size (KDE max ± SD) minimum = 13 ± 6 nm, maximum = 20 ± 12. N = 1365 particles".

The datasets config files are a little more fiddly:

• gaussian filter of (σ = 1) of 1.5 pixels (1–2 nm)
• std masking of 0.7σ for CNDP52, 0.57σ for NDP52-FL, and 1.5σ for the selective masking of terminal domains in NDP52-FL
• min_size of 50 nm2
• Large and small masks were removed by median size of the masked molecules, and then removed objects outside of a customisable size range based on this (no info on what)