Furthermore, the lists of references of retrieved publications were manually checked for additional studies potentially meeting the inclusion criteria and not found by the electronic search.
Before the main analysis, the output from the automatic syllable detection step (as described in the previous section) is manually checked for detection errors.
The results of fragment assignment were manually checked for multiple acyl chain varieties in the same sphingolipid class and reliably detected fragment ions were registered as diagnostic ions.
The obtained reference list of retrieved publications was manually checked for additional references potentially meeting the inclusion criteria and not found by the electronic search.
The references of retrieved publications were manually checked for additional studies that would potentially meet the eligibility criteria and that had not been found by the electronic search.
Partitioning results where manually checked for a large panel of experiments.
EMG activity flagged by this algorithm was also manually checked for false positives.
Sequences were manually checked for ambiguities and assembled in CodonCode Aligner version 1.3.0 (http://www.codoncode.com/).
Gene models are manually checked for errors such as in-frame stops, very short proteins, splits, merges, etc.
In order to rule out possible false positives due to homology artefacts, each variant was manually checked for homology using BLAT search [24].
Array images were manually checked for defects using DChip [20], [21] and then normalized using the RMA algorithm in Affymetrix Expression Console (v1.0).
