Parameterization of disorder predictors for large-scale applications requiring high specificity by using an extended benchmark dataset

Table 1 Summary of the different predictors and their characteristic differences among the methods.

	low sequence complexity	disorder in 3D structures	fully database independent	trained on part of disorder dataset	considers alignment of related sequences
SEG [26, 42]	+	-	+	-	-
CAST [23]	+	-	~^#	-	-
IUPred [25]	-	+	-	-	-
DisEMBL [22]	-	+	-	+	-
DISOPRED2 [24]	-	+	-	+	+^§

^# scores with BLOSUM62
^§ through PSI-BLAST
Different concepts of disorder applied during the development of the respective predictors are indicated by markers: "+" indicates that this concept is very relevant for this predictor, "-" indicates that this concept has not been applied and "~" indicates that this concept is indirectly implied.

ISSN: 1471-2164