I am working with data which contains gaps (from UTrack). A particle may disappear/be lost for a few frames then reappear. In the track data, the missing frames are represented as NaN values. However, vbSPT seems like it is not able to handle this format. I receive the warning dialogue: "the script could not run. Please check that the input data and trajectory field are correctly assigned."
Could vbSPT handle gaps with another format?
Thank you,
Jaime
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi!
vbSPT cannot handle missing positions, unfortunately. It can however handle a large number of trajectories, so one way forward might be to split up your data and analyse the stretches between missing values as distinct, independent trajectories. In that case, it might also be a good idea enforce a lower limit on the trajectory length and discard anything with less than, say 5-10 points (experimetns with simulated data might help you choose).
There, we describe HMM algorithms that handle point-wise localisation uncertainty, and missing values can then be handled by assigning a large localsation uncertainty, say, the on the order of your cell or region of interest.
👍
1
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Another interesting possibility is to try some imputation scheme (see https://en.wikipedia.org/wiki/Imputation_(statistics) ) . This is common in many machine learning applications, but I don't know if anyone has tried it for vbSPT.
👍
1
Last edit: Martin Lindén 2021-12-08
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello,
I am working with data which contains gaps (from UTrack). A particle may disappear/be lost for a few frames then reappear. In the track data, the missing frames are represented as NaN values. However, vbSPT seems like it is not able to handle this format. I receive the warning dialogue: "the script could not run. Please check that the input data and trajectory field are correctly assigned."
Could vbSPT handle gaps with another format?
Thank you,
Jaime
Hi!
vbSPT cannot handle missing positions, unfortunately. It can however handle a large number of trajectories, so one way forward might be to split up your data and analyse the stretches between missing values as distinct, independent trajectories. In that case, it might also be a good idea enforce a lower limit on the trajectory length and discard anything with less than, say 5-10 points (experimetns with simulated data might help you choose).
If you are able to supply a localisation uncertainty to your position trajectories, you might consider some more recent work:
https://github.com/bmelinden/uncertainSPT . Paper: https://doi.org/10.1038/ncomms15115, also discusses how to estimate localisation uncertainty.
https://github.com/bmelinden/vbSPTu . Paper: https://doi.org/10.1016/j.bpj.2018.05.027
There, we describe HMM algorithms that handle point-wise localisation uncertainty, and missing values can then be handled by assigning a large localsation uncertainty, say, the on the order of your cell or region of interest.
Another interesting possibility is to try some imputation scheme (see https://en.wikipedia.org/wiki/Imputation_(statistics) ) . This is common in many machine learning applications, but I don't know if anyone has tried it for vbSPT.
Last edit: Martin Lindén 2021-12-08