Algorithm implementation

The DNetPRO algorithm is made by a sequence of different steps which have to be performed sequentially for a signature extraction. To this purpose, each step can be optimized independently by using the full set of available computational resources¹. In this section we analyze each part of the pipeline, focusing on the optimization strategies used for the algorithm implementation.

The full code is open source and available at [DNetPRO]. The code installation is automatically tested using travis (for Linux and MacOS environments) and appveyor (for Windows environments) at every commit. The installation can be performed using CMake and a full set of installation instructions can be found in the on-line project documentation.

The Python version of the algorithm (see next sections) can be installed via setup.py and the compilable parts of it checked via CMake. The Python installer provides also the full list of dependencies of the project, which will be automatically installed. A full list of example scripts and utilities to obtain the results showed in the next sections can be found in the Github repository. We provide also the complete benchmark pipeline used in our simulations and able to run on cluster environment using the SnakeMake version of it (see next sections).

next »

1. Further optimization can be performed in a cross validation environment and they will be discussed later in this section. ↩