Skip to main content

Table 3 Reverse-engineering verification and validation results for A, B and C

From: Comparative study of three commonly used continuous deterministic methods for modeling gene regulation networks

 

ANN data

SS data

GRLOT data

 

Sparse data set

P inf

0.9620

0.9701

0.8562

0.9308

0.9994

0.9947

0.9546

0.4820

0.6046

P ver

0.9996

0.9797

0.9982

0.9998

0.9988

0.9973

0.9985

0.9941

0.9999

P val

0.9972

0.9870

0.9976

0.9953

0.8635

0.7878

0.9982

0.9927

0.9999

Δ P fit

-0.0024

0.0073

-0.0016

-0.0045

-0.1353

-0.2095

-0.0003

-0.0014

0.0000

 

Detailed data set

P inf

0.9896

0.9992

1.0000

1.0000

0.9999

1.0000

0.6704

0.7070

0.6432

P ver

0.9998

0.9999

1.0000

1.0000

1.0000

1.0000

0.9983

0.9978

0.9987

P val

0.9997

0.9999

1.0000

0.9978

0.9882

0.9891

0.9978

0.9972

0.9996

ΔP fit

-0.0001

0.0000

0.0000

-0.0022

0.0118

0.0109

-0.0005

-0.0006

0.0009

  1. Summary of verification (Step 4) and validation (Step 5) results from reverse-engineering dynamic GRN models from data generated with the reference models A, B and C (Step 1). Here, only same-method results are shown - i.e., ANN model reverse-engineered with ANN method from data generated with ANN reference model, and so on. Results are given for cases A, B, and C for each method (columns below method name from left to right). For the sparse and detailed data sets, P inf and P ver first given for the reverse-engineered models, then P val for the predictions (averaged over 2 values for input increase and decrease). The ΔP fit (for data fitting) is the difference between P ver and P val scores.
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us