This Python program does two things:

1. Compute correlations between aligners' performance and genome complexity. To do this, users must provide (a) the performance of aligners on a set of genomic sequences and (b) the complexity of the sequences. To compute complexity of sequences, use this program: https://github.com/vtphan/sequence-complexity

2. Train and predict aligners' performance based on complexity. Users must also specify the fraction of training set (between 0 and 1). The set of genomic sequences will be used for training and testing.

Dependency

• Python
• Python Scipy

Usage

usage: cross_validation.py [-h]
                           complexity training_portion dir performance_keys
                           [performance_keys ...]

Train and predict short-read alignment performance using different complexity
measures.

positional arguments:
  complexity        file containing complexity values of genomes
  training_portion  fraction of data used for training
  dir               directory containing text files storing aligner
                    performance
  performance_keys  Prec-100, Rec-100, Prec-75, Rec-75, Prec-50, Rec-50, ...

optional arguments:
  -h, --help        show this help message and exit

Example 1: compute correlations of aligners to different measures of complexity.

    python cross_validation.py complexity.txt 1 perf-samples/ Prec-100 Rec-100

Example 2: compute prediction errors of aligners' prediction of alignment accuracy, based on 50% training 50% testing.

    python cross_validation.py complexity.txt 0.5 perf-samples/ Prec-100 Rec-100