###gt-scaffold-assets
Supplementary data and tools to evaluate gt-scaffold
####Data sets
The test data was generated using a Dockerized SGA pipeline (see ./docker). The various input reads were generated from a reference using art_illumina
with the following profile:
art_illumina -sam -i Enterobacteriophage_lambda.fa -l 150 -f 20 -m 400 -s 10 -rs 123 -o e_lambda
=======================================ART======================================
ART_Illumina (2008-2014)
Q Version 2.1.8 (Mar 8, 2014)
Contact: Weichun Huang at whduke@gmail.com
================================================================================
Paired-end sequencing simulation
Total CPU time used: 25.8323
The random seed for the run: 123
Parameters used during run
Read Length: 150
Genome masking 'N' cutoff frequency: 1 in 150
Fold Coverage: 20X
Mean Fragment Length: 400
Standard Deviation: 10
Profile Type: Combined
ID Tag:
Quality Profile(s)
First Read: EMP250R1 (built-in profile)
Second Read: EMP250R2 (built-in profile)