Table 3 Direct comparison of the prediction performance from different combinations of databases and prediction models via FLAME (FLuorophore design Acceleration ModulE)

From: A modular artificial intelligence framework to facilitate fluorophore design

Dataset 1: FluoDB

Object

Algorithms

MAE

MSE

RMSE

λabs

GBRT

13.67

824.24

28.71

SMFluo

21.19

1255.71

35.44

UVVisML

13.94

716.91

26.78

SchNet

22.17

1684.74

41.05

ABT-MPNN

12.66

687.97

26.23

FLSF

12.56

675.34

25.99

Dataset 2: Deep4Chem

λabs

GBRT

24.97

1972.37

44.41

SMFluo

34.2

2992.06

54.7

UVVisML

24.59

1833.93

42.82

SchNet

23.68

2071.75

45.52

ABT-MPNN

22.07

1614.01

40.17

FLSF

24.26

1930.89

43.94

  1. MAE mean absolute error, MSE mean-square error, RMSE root-mean-square error, databsets including our FluoDB and Deep4Chem36; prediction models including previously reported GBRT23, SMFluo29, UVVisML45, SchNet26, ABT-MPNN46, and our FLSF (FLuorescence prediction with fluoroScaFfold-driven model); λabs: maximum absorption wavelength. The unit of MAE/RMSE is nm for λabs.
Back to article page