Skip to main content

Table 1 Average precision/recall scores on the 143 Pfam families from the parameter selection dataset across all four-folds for the three different fragment lengths

From: HIPPI: highly accurate protein family classification with ensembles of HMMs

Method

Fragment Size

 
 

Quarter-length

Half-length

Full-length

 

Prec.

Rec.

Prec.

Rec.

Prec.

Rec.

HIPPI-c (5 %, 0 %)

92.6 %

33.8 %

86.4 %

47.7 %

85.6 %

79.2 %

HIPPI-c (10 %, 0 %)

93.1 %

34.3 %

86.6 %

45.1 %

85.1 %

78.0 %

HIPPI-c (10 %, 40 %)

92.2 %

33.4 %

86.6 %

48.8 %

85.8 %

79.5 %

HIPPI-a (5 %, 0 %)

82.6 %

39.9 %

78.6 %

67.4 %

84.1 %

83.3 %

HIPPI-a (10 %, 0 %)

82.8 %

37.8 %

78.1 %

65.2 %

83.6 %

82.7 %

HIPPI-a (10 %, 40 %)

82.5 %

40.6 %

79.3 %

68.8 %

84.6 %

83.8 %

  1. The methods are labeled as HIPPI-x (X%,Y%), where “-x” is either “-c” (the conservative setting that uses the default gathering cutoff threshold for the inclusion threshold) or “-a” (the aggressive setting that uses no inclusion threshold), X% denotes the maximum decomposition size, expressed as a percentage of the number of sequences in the full dataset, and Y% denotes the minimum average pairwise sequence identity within each subset, expressed as a percentage. The best average precision and recall for the conservative and aggressive versions are bolded