Overview

This repository provides tools and scripts to evaluate catfish, which implements an efficient heuristic for the minimum path flow decomposition problem. Please refer to our paper.

Datasets

We used two datasets to evaluate catfish, namely rnaseq and simulation. They are available at zenodo. The first dataset, rnaseq, contains four sets of perfect splice graphs with different species. The second dataset, simulation, contains two sets of randomly simulated flow graphs with different parameter settings. For both datasets, we provide input instances and ground truth decompositions.

We recommend downloading the data and uncompressed them in the folder of data, so that the scripts in bin can be directly used.

Catcompare

We provide tool catcompare to evaluate the predictions of catfish (or the greedy algorithm) by comparing the decomposition of the algorithm with the ground truth decompositions.

We provide script to compile catcompare (requires automake package):

cd catcompare
./build.sh

This will also link the binary catcompare to bin.

The usage of catcompare is as follows:

catcompare prediction ground-truth > algo.cmp

where the first parameter is the output of catfish (could be also the greedy algorithm implemented within catfish) specified by -o. The second parameter is the ground truth decompositions.

Experiments

We provide scripts in bin to run and analysis the results for the two datasets. You need to link the binary catfish to bin before using these scripts.

1. run.rnaseq.sh: run catfish and the greedy algorithm on rnaseq dataset.
2. run.simulate.sh: run catfish and the greedy algorithm on simulation dataset.
3. compare.rnase.sh: compare the predictions of both methods with the ground truth decompositions for rnaseq dataset.
4. compare.simulation.sh: compare the predictions of both methods with the ground truth decompositions for simulation dataset.
5. stat.rnaseq.sh: collect results for the rnaseq dataset (used for Figures 10--13 and Table 1).
6. stat.simulate.sh: collect results for the rnaseq dataset (used for Table 2 and Table 3).
7. stat.truth.sh: collect statistics for ground-truth decompositions (used for Figure 8 and Figure 9).