Uncertainty in ThunderSTORM results table

[john0...@gmail.com]

Dear Developers and Users,

Can I please check how is the uncertainty [nm] in the ThunderSTORM results table calculated? Is it equal to "FWHM (spot) / SQRT(Number of photon)"?

Thanks,
John

[Martin Ovesný]

Hi John,

it depends. If you use stable version 1.3 then it is always the Thompson-Larson-Webb formula, so your guess is close. If you use the latest daily build version, it get quote complicated. I'm aware that there is no documentation on it yet, because I'm working on my thesis where I explain this and I'll use the text also for documentation when it's finished. The current version goes like this (please be aware this is a working revision of the text...the only reason I publish this prematurely is that there is no documentation other than source code):

Lateral uncertainty:

Axial uncertainty:

Best,

Martin

[Martin Ovesný]

By the way, the missing equation 3.14 is the standard calibration curve as defined in Huang 2008:

In the new versions of ThunderSTORM (the latest daily builds), no matter which calibration model you use, this calibration curve is always estimated, so user doesn't need to worry about what model he used. The reason why we also provide the polynomial model is that it is much simpler for fitting algorithm to estimate the parameters from a stack of beads, since there is not ussually much information. Then it is much better to fit the more complicated model from the polynomial one, because we can generate as many points on the curve as we need for good estimation.

I believe this is all the information necessary to fully answer the question.

Martin

[nah...@brandeis.edu]

Hi Martin,

ThunderStorm is great software, very well designed. I have been having one issue however concerning uncertainty calculations.

The numbers for uncertainty seem much higher than I would predict based on my photons/molecule, sigma values and background signal. I have been trying to recapitulate the uncertainty numbers that ThunderStorm supplies for single localizations given the formula (30 or 31) you supplied in your paper's Supplementary Note. The numbers I get are wildly off what the program is reporting.

Any information or help on this would be greatly appreciated.

Below is an example:

frame x [nm] y [nm] sigma [nm] intensity Offset
1 3939.225465 40123.09245 549.8444414 3421.249853 596.1576773

bkgstd [photon] uncertainty [nm]
2.486917115 15.07335473

Are the formulas above a more recent iteration?

Cheers,
Marc

[Martin Ovesný]

Hi Marc,

yes it's like I wrote above - the more complex equations were introduced in one of the recent daily builds. I can look into this, but I would need you to check whether you still get suspicious results with the new version and if you do, then attach a cropped area around the molecule to your post and let me know your camera settings so I can replicate the problem. It will be just few bytes, so it shouldn't be any problem.

Best,

Martin

[ahei...@umich.edu]

Hi Martin,
Thanks for posting this--I was a little confused about the conflict between the documentation & source code until I found this update. I was just wondering if this updated documentation will be published at some point so it can be cited.
Thanks again,
Abbey

[Martin Ovesný]

Hi Abbey,

it deppends on what you mean by publish. I've just updated the supplementary note (it is attached; or you can navigate to Downloads page at wiki on github - there is a link to our GDrive - go to `paper` directory where the supplementary note for version 1.3+ is). The help built in ThunderSTORM will be updated in couple of hours when compilation finishes (version dev-2015-10-03). As far as any new publications, it's going to be present in my thesis (due end of '15). There will be no extra paper on this topic as I've compiled the formulae from papers where their authors already did the hard part. It depends what you want to cite. I really like the Rieger '14 paper. Although, citing ThunderSTORM would cheer me up too :-) Anyway, you can find all the references in the supplementary note.

Best,

Martin