([HOME]

Statistical analysis of representative 60 pairs: RESULT (c)


(c) Average values of  five metrics

[SUMMARY] (added on 2009-11-22)
  1. Characterization of D2 code, PB code, and DSSP code with respect to dihedral angle patterns

    • D2(86.4%) is the most accurate metric w.r.t. ANGL. (PB is 85.1% and DSSP is 84.5%.)

    • PB(78.2%) is the most sensitive metric w.r.t. ANGL_CORE.  (D2 is 65.6% and DSSP is 63.1%.)
    • The correlation coefficient of PB(43.7) w.r.t ANGL_CORE is better than that of D2(42.9) and that of DSSP(36.8).
    • 90.3% of Non-ANGL_CORE residues are D2-negative. On the other hand, 86.4% are PB-negative and 87.4% are DSSP-negative.

    • D2(95.4%) is the most specific metric w.r.t. ANGL_SUPP. (PB is 91.3% and DSSP is 92.7%.)
    • The correlation coefficient of D2(24.5) w.r.t ANGL_SUPP is better than that of PB(23.9) and that of DSSP(24.1).

  2. Charactrization with respect to the FLEX regions

    • D2(83.1%) and ANGL(83.3%) are more accurate w.r.t. FLEX than others. (PB is 80.1% and DSSP is 81.2%.)

    • ANGL(48.6%) is more sensitive w.r.t. FLEX than others. (D2 is 38.6%, PB is 45.1%, and DSSP is 40.8%.) 
    • D2(91.6%) is more specific w.r.t. FLEX than others. (PB is 86.9%, DSSP is 88.6%, and ANGL is 90.3%.)

    • The correlation coefficient of ANGL(40.2) w.r.t FLEX is better than that of D2(31.1), that of PB(30.1), and that of DSSP(28.8).

  3. Examples (added on 2009-11-25)



[DATA]
(0) Ratio of Variable regions
Variable regions / (Variable regions + Constant regions)  (%)
D2 code PB code DSSP code ANGL code KUZ2 code
Ratio of Variable regions
(Average per chain)		 15.4 19.8 18.0 17.8 6.8
* The table gives the value of the average of the ratio "var(i)/length(i)" of each sequence.
   [NOTE] It is not the ratio of {Sum of var(i)}/{Sum of length(i)}.    (added on 2014-12-09)

(1) Accuracy

 Accuracy = (TP + TN) / (TP + FP + TN + FN)  (%)
D2 code PB code DSSP code ANGL code KUZ2 code
ANGL_CORE 88.7 86.4 86.0 92.4 96.6
ANGL 86.4 85.1 84.5 100.01) 89.0
ANGL_SUPP 57.6 59.7 58.3 63.7 52.7
Flexible regions (FLEX) 83.1 80.1 81.2 83.3 84.8

(2) Sensitivity

 Sensitivity = TP / (TP + FN) (%)
D2 code PB code DSSP code ANGL code KUZ2 code
ANGL_CORE 65.6 78.2 63.1 100.0 64.7
ANGL 52.3 59.9 52.3 100.01) 34.3
ANGL_SUPP 24.1 29.0 26.8 30.3 11.4
Flexible regions (FLEX) 38.6 45.1 40.8 48.6 22.7

(3) Specificity
Specificity = TN / (TN + FP) (%)
D2 code PB code DSSP code ANGL code KUZ2 code
ANGL_CORE 90.3 86.4 87.4 91.1 100.0
ANGL 92.8 88.7 89.9 100.01) 100.0
ANGL_SUPP 95.4 91.3 92.7 100.02) 100.02)
Flexible regions (FLEX) 91.6 86.9 88.6 90.3 97.5

(4) Correlation Coefficient

Correlation Coefficient = (TP·TN - FP·FN) /
sqrt{(TP+FN)(TP+FP)(TN+FP)(TN+FN)} (%)
D2 code PB code DSSP code ANGL code KUZ2 code
ANGL_CORE 42.9 43.7 36.8 67.9 76.7
ANGL 44.9 43.6 40.5 100.01) 52.4
ANGL_SUPP 24.5 23.9 24.1 37.8 20.2
Flexible regions (FLEX) 31.1 30.1 28.8 40.2 30.5

(5) Selectivity

 Selectivity = Sensitivity / (1 - Specificity)
D2 code PB code DSSP code ANGL code KUZ2 code
ANGL_CORE 11.38 8.86 8.31 709.28 (SP=1)
ANGL 12.38 8.58 11.28 (SP=1)1)
 (SP=1)
ANGL_SUPP 8.90 5.98 5.30 (SP=1)
 (SP=1)
Flexible regions (FLEX) 7.18 5.61 5.21 13.27 12.60

    1) identical.
    2) Corrections [Wrong: 99.8 ==> Correct: 100.0]  (2009-11-21)