Table 3 Prediction error of the RF, ULSTM, BLSTM and CNN-LSTM models, as measured by RMSD and std in Å units for TM, ICL, ECL loops of the APO, FA, PIA simulations for 2rh1 and 3p0g states.

From: GPCR molecular dynamics forecasting using recurrent neural networks

Region

MD-RMSD

RF

ULSTM

BLSTM

CNN-LSTM

APO

2rh1

TM \(^{*+}\)

2.0120

0.2660 \(^{**-}\) \({\pm }\) 0.0040

0.1390 \({\pm }\) 0.0080

0.1460 \({\pm }\) 0.0080

0.1430 \({\pm }\) 0.0020

ICL

2.3200

0.3150 \(^{**-}\) \({\pm }\) 0.0090

0.1790 \({\pm }\) 0.0090

0.1910 \({\pm }\) 0.0070

0.1830 \({\pm }\) 0.0019

ECL

2.0800

0.2770 \(^{**-}\) \({\pm }\) 0.0020

0.1810 \({\pm }\) 0.0090

0.1850 \({\pm }\) 0.0080

0.1880 \({\pm }\) 0.0020

APO

3p0g

TM \(^{**+}\)

1.8652

0.1746 \(^{**-}\) \({\pm }\) 0.0019

0.1438 \({\pm }\) 0.0052

0.1444 \({\pm }\) 0.0041

0.1387 \({\pm }\) 0.0012

ICL

1.5981

0.2213 \(^{**-}\) \({\pm }\) 0.0027

0.1887 \({\pm }\) 0.0056

0.1920 \({\pm }\) 0.0056

0.1878 \({\pm }\) 0.0019

ECL

1.5306

0.2101 \(^{**-}\) \({\pm }\) 0.0010

0.1781 \({\pm }\) 0.0054

0.1791 \({\pm }\) 0.0042

0.1725 \({\pm }\) 0.0011

FA

2rh1

TM \(^{**+}\)

1.1757

0.1672 \(^{*-}\) \({\pm }\) 0.0021

0.1350 \({\pm }\) 0.0042

0.1358 \({\pm }\) 0.0041

0.1349 \({\pm }\) 0.0042

ICL

1.1518

0.2533 \(^{**-}\) \({\pm }\) 0.0033

0.1823 \({\pm }\) 0.0041

0.1850 \({\pm }\) 0.0041

0.1856 \({\pm }\) 0.0043

ECL

1.1476

0.2157 \(^{**-}\) \({\pm }\) 0.0038

0.1713 \({\pm }\) 0.0056

0.1705 \({\pm }\) 0.0054

0.1716 \({\pm }\) 0.0050

FA

3p0g

TM \(^{**+}\)

1.7808

0.1736 \(^{**-}\) \({\pm }\) 0.0021

0.1362 \({\pm }\) 0.0052

0.1343 \({\pm }\) 0.0036

0.1230 \({\pm }\) 0.0023

ICL

1.4829

0.2132 \(^{**-}\) \({\pm }\) 0.0022

0.1884 \({\pm }\) 0.0059

0.1863 \({\pm }\) 0.0045

0.1834 \({\pm }\) 0.0031

ECL

1.4983

0.2071 \(^{**-}\) \({\pm }\) 0.0024

0.1744 \({\pm }\) 0.0066

0.1737 \({\pm }\) 0.0047

0.1682 \({\pm }\) 0.0029

PIA

2rh1

TM \(^{**+}\)

1.7686

0.1782 \(^{**-}\) \({\pm }\) 0.0043

0.1388 \({\pm }\) 0.0029

0.1388 \({\pm }\) 0.0028

0.1338 \({\pm }\) 0.0066

ICL

1.5180

0.2712 \(^{**-}\) \({\pm }\) 0.0073

0.1896 \({\pm }\) 0.0026

0.1889 \({\pm }\) 0.0046

0.1850 \({\pm }\) 0.0092

ECL

1.4525

0.2134 \(^{**-}\) \({\pm }\) 0.0027

0.1725 \({\pm }\) 0.0035

0.1715 \({\pm }\) 0.0032

0.1675 \({\pm }\) 0.0061

PIA

3p0g

TM \(^{**+}\)

1.7853

0.1888 \(^{**-}\) \({\pm }\) 0.0047

0.1365 \({\pm }\) 0.0071

0.1298 \({\pm }\) 0.0053

0.1324 \({\pm }\) 0.0016

ICL

1.4847

0.2260 \(^{**-}\) \({\pm }\) 0.0048

0.1921 \({\pm }\) 0.0075

0.1856 \({\pm }\) 0.0065

0.1890 \({\pm }\) 0.0018

ECL

1.4513

0.2188 \(^{**-}\) \({\pm }\) 0.0029

0.1675 \({\pm }\) 0.0082

0.1591 \({\pm }\) 0.0063

0.1623 \({\pm }\) 0.0020

Mean

TM \(^{**+}\)

1.7313

0.1914 \(^{**-}\) \({\pm }\) 0.0032

0.1382 \({\pm }\) 0.0054

0.1382 \({\pm }\) 0.0046

0.1343 \({\pm }\) 0.0003

ICL

1.5926

0.0025 \(^{**-}\) \({\pm }\) 0.0049

0.1867 \({\pm }\) 0.0058

0.1881 \({\pm }\) 0.0054

0.1856 \({\pm }\) 0.0037

ECL

1.5267

0.2237 \(^{**-}\) \({\pm }\) 0.0025

0.1741 \({\pm }\) 0.0064

0.1731 \({\pm }\) 0.0053

0.1717 \({\pm }\) 0.0032

  1. Statistics of significantly worse\(^{(-)}\) and significantly better\(^{(+)}\) differences with p \(< 0.05^{*}\) and p\(< 0.01^{**}\) are included. The symbols in the first column indicate significantly better differences between regions, while in the model results columns indicate significantly worse differences between models. The Mean row at the bottom of the table indicates mean errors across regions. Bold values show the minimum error accross regions and models.
Back to article page