Howdy all,
I've been working on a support-vector machine based variant caller, and it's finally at the point where others may find it useful. It calls variants from .BAM or .SAM files in a manner similar to samtools / mpileup or the GATK, and does so with greater accuracy than most other variants callers.
Here's a few quick facts:
Source code and downloads available from github
Feel free to check out an early version of the manuscript
One requirement : It depends on libsvm. (libsvm is also available in many linux repositories, try your package manager first)
Happy SNP calling.
I've been working on a support-vector machine based variant caller, and it's finally at the point where others may find it useful. It calls variants from .BAM or .SAM files in a manner similar to samtools / mpileup or the GATK, and does so with greater accuracy than most other variants callers.
Here's a few quick facts:
- More accurate than GATK, samtools, or FreeBayes
- SNPS only right now
- Parallelizes easily
- Uses a support-vector machine to separate true from false positive variants
- About as fast as the GATK, very low memory requirements
- Can 'learn' from additional training data and become more accurate over time
Source code and downloads available from github
Feel free to check out an early version of the manuscript
One requirement : It depends on libsvm. (libsvm is also available in many linux repositories, try your package manager first)
Happy SNP calling.
Comment