Hidden-Markury

FRET Trace Analysis with Hidden-Markov-Model (Jupyter Notebook)

Installation

1. Install a C compiler for hmmlearn (optional, if missing)

The basic implementation of the Hidden-Markov-Process and the model optimization is based on build-in functions of the python library hmmlearn[1]. hmmlearn requires a C compiler. If you don't have a C compiler installed, the installation of hmmlearn will raise an error.

On Windows:

• Go to https://visualstudio.microsoft.com/downloads/ -> Select "Tools for Visual Studio 2019" -> "Build Tools for Visual Studio 2019" -> Download
• Install "Python development" with "Python native development tools" selected

On Linux/Mac:

• Not tested yet! Installation of hmmlearn (see next section) should raise informative error what to do.

2. Create new environment (optional)

Create a new python environment to install the required packages

Example in Anaconda prompt: conda create -n hmm

3. Install hmmlearn

• pip install --upgrade --user hmmlearn

If pip installation fails, you can try to directly download and install the wheel.

• (alternative: download wheel from http://pypi.fcio.net/simple/hmmlearn/ and install with pip install PATH_TO_WHEEL)

4. Install other required packages

The following packages are needed for full functionality:

• numpy
• pandas
• matplotlib
• scipy

Test system

• Windows 10, 64 bit system
• Anaconda3 (64 bit)
• Python 3.7 environment
• Packages:
  • hmmlearn 0.2.2
  • numpy 1.17.3
  • pandas 0.25.3
  • matplotlib 3.1.1
  • scipy 1.3.1

Getting Started

1. Clone or download this repository

Clone repository via git

git clone https://github.com/ChristianGebhardt/Hidden-Markury

or clone/download repository manually:

2. Open "Hidden-Markury Notebook" in jupyter notebook

• Activate python environment in python console (e.g. activate hmm in Anaconda prompt) with required packages installed (see Installation)
• Navigate into repository folder cd PATH_TO_HIDDEN-MARCURY
• Start jupyter with command jupyter notebook
• Run notebook "Hidden-Marcury Notebook.ipynb"

3. Check installations

There is an extra import file, which checks and imports all the required packages at the beginning. If everything works fine, it should look the following: Otherwise, an error should inform you what to do.

4. Run example analysis

There are several example traces provided in the folder examples provided by the "kinSoftChallenge"[2]. Those traces can directly be used to get familiar with the software.

If you run the notebook cell-by-cell and follow the instructions, you should get the optimized model with the fitted transition rates and the state prediction for the analyzed traces:

Author(s)

• Christian Gebhardt - Initial work - ChristianGebhardt

See also the list of contributors who participated in this project.

License

This project is licensed under the MIT License - see the LICENSE file for details

Release Notes

Version

v1.0.1 - 2020-02-25

Features

• Global trace analysis
• 1D (E-Trace) or 2D (photon channels) analysis
• Model optimization
• Lifetime fitting
• Sample variation (error estimation)

Acknowledgments

This software was designed and created by Christian Gebhardt within ongoing work of the Cordes lab (http://www.mikrobiologie.biologie.uni-muenchen.de/forschung/ag_cordes/index.html). The project was financed by the German Science Foundation (SFB863, TP A13, to Thorben Cordes), an ERC Starting Grant (No. 638536 – SM-IMPORT to Thorben Cordes) and a PhD fellowship of the Studienstiftung des deutschen Volkes (to Christian Gebhardt).

Bibliography

[1] https://hmmlearn.readthedocs.io/en/latest/index.html

[2] https://sites.google.com/view/kinsoftchallenge/