Table 1 Data on the analytical methods reported based on QSRR

1

QSRR for improving structural identification in non-targeted metabolomics

ChromSword—to build the QSRR model,

ACD/ChromGenius—descriptors calculation

34 groups of compounds (248 compounds in total)

HPLC/MS

Agilent 1100 HPLC

QTOF-2 mass spectrometer

PLS (Partial least squares)

R² = 0.8512

%RMSE = 8.45

SD = 46

N = 191

[101]

2

QSRR of ionic liquid cations for characterization of stationary phase for HPLC

Hyperchem 7.5 package with Chemplus—for molecular modelling

8 ionic liquid cations

HPLC-Shimadzu-LC-10Avp pump with manual injector

MLR (Multiple linear regression),

PCA (principal component analysis)

Log k versus logP

R² =  > 0.9 for 95% confidence intervals. (For 6 different columns)

PCs- 97.3% total data variance

[102]

3

QSRR for HPLC Rt prediction of small molecules: endogenous metabolite

Gaussian 03—3D structure calculation,

Dragon software version 5.5—for descriptor calculation

9 classes of banned drugs, metabolites, and salts (totally of 146 analytes)

LC–MS method

MLR (Multiple linear regression),

PLS (Partial least squares)

R = 0.97

R² = 0.95

RMSE = 3.45

F = 778

[103]

4

QSRR of peptides behaviour in RPLC

MATLAB 7.0.1,

Dragon software 5.0,

Hyperchem 7.5 package

Training set 50 and test set 19, totally of 69 peptides were used

Four different columns with Merck-Hitachi LaChrom system

GA (Genetic algorithm) to MLR (Multiple linear regression)

R² = 0.62–0.86

RMSE = 1.99–2.99

[104]

5

QSRR analysis of antihypertensives and diuretics

Chempro, Chemaxon, Marvin 4.0.5,

Gaussian 98

Totally 17 compounds

SpectraPhoresis 500-

Capillary electrophoretic system equipped with UV detector

Optimized ligands and ligand—BCD complexes

R² = 0.914

%RMSE = 0.092

F ratio = 45.897

[105]

6

QSRR modelling of morphine and its derivatives

Hyperchem 8.0,

VLifeMDS 4.4—descriptors calculation,

Molegro data modeller

Training set 42 and test set 15, totally 57 morphine and its derivatives

GC–LC method

ANN (Artificial neural network)

R² = 0.9559

%RMSE = 0.399

[106]

7

QSRR modelling of coffee flavour compounds

Hyperchem,

CODESSA—descriptor calculation

Training set 36 and test set16, totally 52 compounds

NA

SVM (Support vector machines)

Satisfactory relationship between the molecular descriptor and retention time of coffee

[107]

8

QSRR for chromatographic behaviour prediction of antiviral drugs

Hyperchem 8.0, ACD/ChemSketch, QSRR Automator, ADMETlab

Training set 14 and test set 5, totally 19 antiviral drugs

Agilent LC-1200 system. (RP-HPLC Method)

MLR (Multiple Linear Regression),

SVR (Support vector for regression)

R² = 0.881

RMSE = 0.640

[108]

9

QSRR for naturally occurring phenolic compounds

DRAGAN version 3,

MOPAC2009—descriptor calculation

Training set 30 and test set 9, totally 39 phenolic compounds

RPLC-MS method

ANN (Artificial neural network),

MLR (Multiple linear regression)

R² > 0.80

N = 30

[109]

10

QSRR for prediction of pesticide retention time

ACD software,

ChemDes—descriptor calculation

Total 843 pesticides

UHPLC method

MLR (Multiple linear regression) and SVM (Support vector machines)

R² = 0.85

RMSE = 0.69

[110]

11

QSRR modelling of selected antipsychotics and their impurities

Chemaxon,

MOPAC/AMI of chem3D ultra-7.0.0

The training set 147, test set 31, and validation set 32, totally 210 compounds

Thermo scientific Dionex ultra-3000

(RP-HPLC method)

ANN (Artificial neural network)

Training set

R² = 0.9962

RMSE = 0.2954

Test set

R² = 0.9829

RMSE = 0.3633

Validation set

R² = 0.9927

RMSE = 0.4864

[111]

12

QSRR for mycotoxins and fungal metabolites

ISIS Draw-2.5,

HyperChem,

CODESSA2.63—descriptor calculation

Total 83 compounds

LC–MS/MS method

RBFNN (Radial basis function neural network)

Training set

R² = 0.8933

RMSE = 0.9406

Test set

R² = 0.8709

RMSE = 1.2892

[112]