Massively parallel sequencing has tremendous diagnostic potential but requires enriched templates for sequencing. Here we report the validation of an array-based sequence capture method in genetically heterogeneous disorders. The model disorder selected was AR ataxia, using five subjects with known mutations and two unaffected controls. The genomic sequences of seven disease genes, together with two control loci were targeted on a 2-Mb sequence-capture array. After enrichment, the patients' DNA samples were analyzed using one-quarter Roche GS FLX Titanium sequencing run, resulting in an average of 65 Mb of sequence data per patient. This was sufficient for an average 25-fold coverage/base in all targeted regions. Enrichment showed high specificity; on average, 80% of uniquely mapped reads were on target. Importantly, this approach enabled automated detection of deletions and hetero- and homozygous point mutations for 6/7 mutant alleles, and greater than 99% accuracy for known SNP variants. Our results also clearly show reduced coverage for sequences in repeat-rich regions, which significantly impacts the reliable detection of genomic variants. Based on these findings we recommend a minimal coverage of 15-fold for diagnostic implementation of this technology. We conclude that massive parallel sequencing of enriched samples enables personalized diagnosis of heterogeneous genetic disorders and qualifies for rapid diagnostic implementation.