A set of algorithms for protein identification using a sub-nanopore.

The package is implemented in Python 2.7 and requires no installation. However, it depends on some third party Python packages:

• numpy [http://www.numpy.org/]
• scipy [http://www.scipy.org/]
• scikit-learn [http://scikit-learn.org/]
• matplotlib [http://matplotlib.org/]

These packages should be instaled in your system. You can use either system package manager (e.g. apt-get in Ubuntu) or pip for installation:

pip install package_name

To download the datasets used in the paper, type:

wget https://github.com/fenderglass/datasets/raw/master/nano-align/nano-align.zip

The package also contains the trained Random Forest and SVR model for convenience.

This is a quick example of Nano-Align pipeline, assuming that you have downloaded the manuscript datasets into 'data' directory. Please see the detailed usage information below.

Train Random Forest model on H32 nanospectra:

./train-model.py rf data/nanospectra/H32.mat h32_rf.pcl

Perform identification of H4 nanospectra:

./identify.py data/nanospectra/H4.mat h32_rf.pcl

Plot H32 nanospectra against the RF and MV models:

./plotting/models-fit.py data/nanospectra/H32.mat h32_rf.pcl,-

Plot identification p-values as a function of number of nanospectra in a cluster:

./plotting/identification-pvalues.py data/nanospectra/H4.mat h32_rf.pcl

See detailed description of the parameters for each script by specifying "-h" option.

Trains the regression model based on Random Forest / SVR, given nanospectra of a known protein. The output file (model) then is then used as an input for other algorithms.

Performs protein identification and estimates p-values. It takes trained RF/SVR model as an input.

There is a number of scripts that can be used to visualize different features of the data. They are located in the "plotting" directory

Plots nanospectra against the corresponding regression models.

Plots identification p-values depending on the cluster size

Plots the frequency distribution of a multiple sets of nanospectra, originating from different proteins.

Plots volume- or hydrophilicity-related bias of the model.

The scripts located in "scripts" directory provide some extra functionality for the data preprocessing and analysis.

Creates a protein database with the certain protein lengths from a bigger FASTA database (such as human proteome).

Given blockades singals and trained SVR model, for each blockade determines if was recorded from straight (N- to C-terminus) or reverse protein translocation. Reverse translocations are then flipped.

Merges multiple .mat files into one.

Adds protein sequence labels to .mat file - a prerequisite for the further analysis.