Table 2.
Important chemical features for classification models. Top three physicochemical features for viral targets where the models classified chemicals as active vs inactive relative to broad inhibition or activition rather than a specific assay value (e.g. Ki, IC50, and AC50).
| Feature | Target | Description |
|---|---|---|
| Mor18s | BRD4 | signal 18/weighted by I-state |
| SpMAD_G/D | BRD4 | spectral mean absolute deviation from distance/distance matrix |
| SpMax3_Bh(p) | BRD4 | largest eigenvalue n. 3 of Burden matrix weighted by polarizability |
| P_VSA_LogP_3 | HDAC2 | P_VSA-like on LogP, bin 3 |
| SHED_DA | HDAC2 | SHED Donor-Acceptor |
| SHED_DL | HDAC2 | SHED Donor-Lipophilic |
| G(N..N) | IDE | sum of geometrical distances between N..N |
| SM1_Dz(i) | IDE | spectral moment of order 1 from Barysz matrix weighted by ionization potential |
| Wap | IDE | all-path Wiener index |
| CATS2D_08_DA | TBK1 | CATS2D Donor-Acceptor at lag 08 |
| F08[N–N] | TBK1 | Frequency of N - N at topological distance 8 |
| P_VSA_e_3 | TBK1 | P_VSA-like on Sanderson electronegativity, bin 3 |
| H7m | PRKACA | H autocorrelation of lag 7/weighted by mass |
| H7s | PRKACA | H autocorrelation of lag 7/weighted by I-state |
| RDF060m | PRKACA | Radial Distribution Function - 060/weighted by mass |
| GATS6e | MARK3 | Geary autocorrelation of lag 6 weighted by Sanderson electronegativity |
| GATS6m | MARK3 | Geary autocorrelation of lag 6 weighted by mass |
| Mor02m | MARK3 | signal 02/weighted by mass |
| CATS2D_02_DL | IMPDH2 | CATS2D Donor-Lipophilic at lag 02 |
| CATS3D_07_DL | IMPDH2 | CATS3D Donor-Lipophilic BIN 07 (7.000–8.000 Å) |
| NaasC | IMPDH2 | Number of atoms of type aasC |
| C-039 | ABCC1 | Ar-C(=X)-R |
| VE2sign_Dz(p) | ABCC1 | average coefficient of the last eigenvector from Barysz matrix weighted by polarizability |
| VE3sign_Dz(v) | ABCC1 | logarithmic coefficient sum of the last eigenvector from Barysz matrix weighted by van der Waals volume |
| Mor31s | ABHD12 | signal 31/weighted by I-state |
| RTi+ | ABHD12 | R maximal index/weighted by ionization potential |
| VE3sign_Dz(p) | ABHD12 | logarithmic coefficient sum of the last eigenvector from Barysz matrix weighted by polarizability |
| E2m | BRD2 | 2nd component accessibility directional WHIM index/weighted by mass |
| GATS2m | BRD2 | Geary autocorrelation of lag 2 weighted by mass |
| TDB03i | BRD2 | 3D Topological distance based descriptors - lag 3 weighted by ionization potential |
| MAXDP | COMT | maximal electrotopological positive variation |
| nDB | COMT | number of double bonds |
| P_VSA_MR_2 | COMT | P_VSA-like on Molar Refractivity, bin 2 |
| CATS2D_02_AL | DNMT1 | CATS2D Acceptor-Lipophilic at lag 02 |
| Mor04s | DNMT1 | signal 04/weighted by I-state |
| VE3sign_Dt | DNMT1 | logarithmic coefficient sum of the last eigenvector from detour matrix |
| ChiA_B(i) | EIF4H | average Randic-like index from Burden matrix weighted by ionization potential |
| F05[C–O] | EIF4H | Frequency of C - O at topological distance 5 |
| NaasC | EIF4H | Number of atoms of type aasC |
| CENT | LOX | centralization |
| EE_G | LOX | Estrada-like index (log function) from geometrical matrix |
| VE2_D/Dt | LOX | average coefficient of the last eigenvector (absolute values) from distance/detour matrix |
| Eta_D_beta | MARK2 | eta measure of electronic features |
| Mor29v | MARK2 | signal 29/weighted by van der Waals volume |
| SpPosA_B(i) | MARK2 | normalized spectral positive sum from Burden matrix weighted by ionization potential |
| CATS2D_07_AL | NEK9 | CATS2D Acceptor-Lipophilic at lag 07 |
| CATS2D_08_AL | NEK9 | CATS2D Acceptor-Lipophilic at lag 08 |
| TDB05p | NEK9 | 3D Topological distance based descriptors - lag 5 weighted by polarizability |
| CATS2D_06_DL | NEU1 | CATS2D Donor-Lipophilic at lag 06 |
| TDB04i | NEU1 | 3D Topological distance based descriptors - lag 4 weighted by ionization potential |
| X3A | NEU1 | average connectivity index of order 3 |
| nR06 | RHOA | number of 6-membered rings |
| R8s+ | RHOA | R maximal autocorrelation of lag 8/weighted by I-state |
| SpMin1_Bh(m) | RHOA | smallest eigenvalue n. 1 of Burden matrix weighted by mass |
| CATS3D_08_NL | SIRT5 | CATS3D Negative-Lipophilic BIN 08 (8.000–9.000 Å) |
| O-057 | SIRT5 | phenol, enol, carboxyl OH |
| SpMax2_Bh(s) | SIRT5 | largest eigenvalue n. 2 of Burden matrix weighted by I-state |
| CATS2D_04_AL | TK2 | CATS2D Acceptor-Lipophilic at lag 04 |
| JGI3 | TK2 | mean topological charge index of order 3 |
| MATS1i | TK2 | Moran autocorrelation of lag 1 weighted by ionization potential |
| P_VSA_e_3 | VCP | P_VSA-like on Sanderson electronegativity, bin 3 |
| RDF020p | VCP | Radial Distribution Function - 020/weighted by polarizability |
| SpMaxA_AEA(dm) | VCP | normalized leading eigenvalue from augmented edge adjacency mat. weighted by dipole moment |