Quantification of single-molecule FRET trajectories containing non-instantaneous transitions

Slopey was written to handle smFRET data that consist of relatively long-lived “flat bits,” where the intensity values appear to be Gaussian scatter around a mean value (that is, they seem to be well described as a constant intensity value with some noise), separated by “slopey” bits, which, though fast, are not instantaneous jumps. Current HMM fitting techniques in the field assume Gaussian emissions with instantaneous transitions. An introduction to Slopey and its application to the quantification of timeseries data for a particular enzyme can be found here.

Building

python setup.py build_ext --inplace

The makefile will take care of this for you unless you have some crazy directory structure with spaces.

Hard-coded limits on numbers of slopey bits and frames

The cython file slopey/fast.pyx has hard-coded limits on the number of slopey bits (NUM_SLOPEY_MAX) and the number of frames of data (NUM_FRAMES_MAX). That is, those limits are set at compile time, but can be increased by editing slopey/fast.pyx and changing these lines (which may not be next to each other):

cdef int NUM_SLOPEY_MAX = 25
cdef int NUM_FRAMES_MAX = 50000

There's essentially no limit to how large you set these, but for ease of implementation they're preset.

Running Steph’s Matlab code

Make a copy of the sample global params.yml file, edit the parameters as needed for your data, and save in the same directory as the .mat files to be analyzed (called <datadir> below).

Each .mat file can also be analyzed with its own parameters, if a .yml with the same filename but with extension .params.yml instead of .mat is present. To use the information in the goodtraces.txt parameters file from pyhsmm (see Traces) to automatically generate trace-specific parameters, first run

ConvertGoodTracesToYAML(<datadir>)

Because Steph does have a crazy directory structure with spaces, you will need to run the build command as above. If you try to run slopey either through the Terminal or by running RunSlopeyAnalysis in Matlab, and get:

Compiling low-level code
python: can't open file 'setup.py': [Errno 2] No such file or directory

then re-build.

With a crazy directory structure with spaces like Steph's, you will also need to create a symbolic link to each data directory. In Terminal, run:

cd “~/Documents/UCSF/.../Symlinks_Data” 
ln -s “/Users/Steph/Documents/.../smFRET data analysis/<datadir>” DataDirName 

To run slopey, either run in the Terminal:

./Analyze_Slopey.sh Symlinks_Data/DataDirName

with optional additional arguments: clean, -dr (prints debug info from make), -j4 (runs multi-threaded; the default in RunSlopeyAnalysis is -j2). Or, run in Matlab:

RunSlopeyAnalysis(DataDirName)

If you get:

make: *** [results_slopey/*_Results.results.pkl] Abort trap: 6

this is likely a trace with one or more long dwells to which pyhsmm assigned many fast-switching states. If pyhsmm is used for initialization, you’ll have problems.

Fix: edit the *_Results.params.yml file (or create a new one) such that

translocation_frame_guesses: [x1,x2,x3]

has reasonable values for x1, x2, x3, etc.

Once slopey is running successfully, it will print out accept proportions for each trace. Optimally they should be ~0.2.

Cropping: Crop traces as little as possible. The more information slopey has about each flat bit's (noisy) distribution of intensity values, the better it will find the intervening slopey bits. This is somewhat in contrast to the discrete-time HMM based on pyhsmm (see Traces and pyhsmm), where long noisy flat bits are likely to be assigned multiple states.

long-term possible todos

• [ ] learn the noise parameters
• [ ] infer number of slopey bits (RJ MCMC)
• [ ] 'energy ladders' in the prior