BioSSA — Tool For Analysis of Gene Expression

Aim of the project is to create the package, which is designed for analysis of 2D data or 3D surface irregular data in cylindrical projection. Interface of the package should be convenient for processing of data of gene expression of drosophila fly.

  • Input data: points \((x_i, y_i, f_i)\) given on a planar surface.
  • Output: decomposition of data on trend (pattern) and noise, construction of the noise model.

Watch GitHub Repository

Package features

  1. Loading data from files given in the following formats (x,y, gene expression).
  2. Decomposition of data
    • Transformation of data to the scale 0–100% by AP and DV coordinates
    • Choice of rectangular area for analysis
    • Regularization of data (interpolation to regular grid), the step of the grid can be set.
    • Decomposition of regularized data by the 2D-SSA method
    • Interpolation of data to initial coordinates.
  3. Checking the decomposition quality
    • Profiles of patterns along AP or DV axis
    • Profiles of noise along AP or DV axis
  4. Analysis of residuals, model of residuals
    • Graphs of dependence of noise on patters,
    • Estimation of noise model: additive, multiplicative or Poissonian model.
  5. Pictures with the results of decomposition in 2D or 3D forms.

Installation

The package is implemented as an R-package.

Latest binary build for Windows could be found here.

It essentially uses the Rssa package. You can also watch Rssa GitHub repository.

Also it uses multidimensional spatial interpolation procedures from our LinearInterpolator package, which requires the CGAL library to be installed. For Windows CGAL compilation is rather complicated, so it’s better to use precompiled binary.

You can install both LinearInterpolator and BioSSA packages from GitHub directly (using Hadley’s package devtools) by the following code:

install.packages("devtools")
library(devtools)
install_github("BioSSA/LinearInterpolator")
install_github("BioSSA/BioSSA")

Be careful, devtools package has some additional dependences.

Method

Decomposition of the image into a sum of pattern and noise is fulfilled by Shaped 2D/3D Singular Spectrum Analysis.

Support

This work is supported by the NG13-083 grant of Dynasty Foundation.

This work is supported by the RFBR grant 16-04-00821.