In molecular biology, gene prioritization is the task of identifying the most promising genes to submit to experimental studies. It is an important step to narrow down the target of wet-lab analyses such as validations and screenings, in order to optimize their results.

Scuba is a computational approach for gene prioritization, based on the integration of gene functional associations. In particular, it focuses on the identification of novel putative associations between genes and genetic disorders.

Let us consider the following entities:

• a group of genes known to be related to a disorder
• a group of candidate genes to be prioritized, that can even be the whole set of human genes
• a collection of functional associations among genes, deriving from multiple data sources The goal of Scuba is to prioritize the group of candidate genes by exploiting their functional connections with the group of known genes. The output is a rank of the candidate genes, ordered from the most to the least promising.

Running this code may be a computationally intensive procedure. Validations mentioned in the documentation were performed on a computer cluster.

This code was tested using Python 2.7.14, Numpy 1.14.2, Pandas 0.23.4 and Scikit-learn 0.19.1.

Scuba is usable from command line by typing python launcher.py [arguments]. For information on the arguments, type python launcher.py --help or see the documentation in launcher.py.

Folder Toy data contains simple inputs and corresponding outputs that can be used to check code functioning.

Script compute_kernel.py allows creating kernel matrices from available datasets through a few kernel functions. For instance, starting from a gene network adjacency matrix such as dataset1_adjacencyMatrix.txt, run:

python compute_kernel.py --input-matrix <path to file>/dataset1_adjacencyMatrix.txt --kernel-function MDK --output <path to file>/dataset1_Matrix_MDK3 -p 3

to get dataset1_Matrix_MDK3.npy. Analogously, for a numerical feature matrix such as dataset2_featureMatrix.txt the RBF function can be used.

To prioritize the whole set of genes represented by the kernel matrices, run:

python launcher.py --kernel-list <path to file>/my_kernels.txt --disease-genes <path to file>/my_disease_genes.txt --output <path to file>/my_output

or, optionally, if a list of candidate genes is available:

python launcher.py --kernel-list <path to file>/my_kernels.txt --disease-genes <path to file>/my_disease_genes.txt --output <path to file>/my_output --candidate-genes <path to file>/my_candidates.txt

Alternatively, to perform the so-called "unbiased" (CAFA-like) validation described in the paper, run:

python launcher.py --kernel-list <path to file>/my_kernels.txt --disease-genes <path to file>/seeds_unbiased_Ensembl.txt --candidate-genes <path to file>/candidates_unbiased_Ensembl.txt --output <path to file>/my_output -f 4 -i 11 --test unbiased

Parameters f and i (and model-selection) can be changed as needed, however note that f should not be greater than 4 as there is a test case with only 4 known disease genes.

If you use Scuba, please cite: Zampieri, G., Van Tran, D., Donini, M., Navarin, N., Aiolli, F., Sperduti, A., & Valle, G. (2018). Scuba: scalable kernel-based gene prioritization. BMC bioinformatics, 19(1), 23.

Scuba - Gene prioritization.
Copyright (C) 2016 by Guido Zampieri, Dinh Tran, Michele Donini.

Scuba is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.

Scuba is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License along
with Scuba. If not, see http://www.gnu.org/licenses/.