NGS is frequently used to identify mutations in DNA samples from individual patients or experimental organisms. Sequencing can be done at the whole-genome scale; RNA-seq, which targets expressed genes; exome capture, which targets specific exon regions captured by hybridization to probes of known sequence; or amplicons for genes or regions of interest. In all cases, sequence variants are detected by alignment of NGS reads to a reference sequence and then identification of differences between the reads and the reference. Variant detection algorithms must distinguish between random sequencing errors, differences caused by incorrect alignment, and true variants in the genome of the target organism. Various combinations of base quality scores, alignment quality scores, depth of coverage, variant allele frequency, and the presence of nearby sequence variants and indels are all used to differentiate true variants from false positives. Recent algorithms have also made use of machine learning methods based on training sets of genotype data or large sets of samples from different patients/organisms that are sequenced in parallel with the same sample preparation methods on the same NGS machines.