Ligand-Based Virtual Screening Based on the Graph Edit Distance

. 2021 Nov 25;22(23):12751. doi: 10.3390/ijms222312751

Algorithm 1 Costs learning.

Input (Learning Set, Initial edit costs,

M a x_I t e r

)
Output (Learnt edit costs)

Initialise:

$i t e r = 1$ .

$C_{1}$ , …, $C_{n}$ = Initial edit costs.

While $i t e r \leq M a x_I t e r$ :
Classify all molecules with nearest neighbour and $G E D$ (Equation (1)) using $C_{1}, \dots, C_{n}$ .
Compute $D_{j}$ and $D_{j}^{'}$ : (Equations (2) and (3)) for all $G_{j}$ incorrectly classified.
Deduce $G_{i}$ (Equation (4)).
Compute $α_{t}$ , $t = 1, \dots, m$ and $α_{t}^{'}$ , $t = 1, \dots, m^{'}$ : (Equation (12)).
Compute ${\bar{C}}_{1}$ , …, ${\bar{C}}_{n}$ (Equation (13)).
Update costs: $C_{t} = {\bar{C}}_{t}, t = 1, \dots, n$ .
$i t e r = i t e r + 1$ .

End While

End Algorithm