Skip to main content

Table 1 The prompts we use in this work. [FG], [FG1], and [FG2] refer to any functional group, and [VALUE] refers to a real number

From: Instruction multi-constraint molecular generation using a teacher-student large language model

Task

Prompt

2FG

The molecule contains [FG1],[FG2]

FG+LogP

The molecule contains [FG]. Its LogP is [VALUE]

FG+QED

The molecule contains [FG]. It has a high QED score

FG+SA

The molecule contains [FG]. It has good synthetic accessibility

FG+DRD2

The molecule contains [FG]. It is active to DRD2

FG+GSK3

The molecule contains [FG]. It is active to GSK3

FG+BBB

The molecule contains [FG]. It can pass through the blood-brain barrier

FG+HIA

The molecule contains [FG]. It can be absorbed by human intestinal

FG+DRD2+QED

The molecule contains [FG]. It is active to DRD2. It has a high QED score

FG+GSK3+QED

The molecule contains [FG]. It is active to GSK3. It has a high QED score

FG+BBB+QED

The molecule contains [FG]. It can pass through the blood-brain barrier. It has a high QED score

FG+HIA+QED

The molecule contains [FG]. It can be absorbed by human intestinal. It has a high QED score

FG+DRD2+QED+SAs

The molecule contains [FG]. It is active to DRD2. It has a high QED score. It has good synthetic accessibility

FG+GSK3+QED+SAs

The molecule contains [FG]. It is active to GSK3. It has a high QED score. It has good synthetic accessibility

FG+BBB+QED+SAs

The molecule contains [FG]. It can pass through the blood-brain barrier. It has a high QED score. It has good synthetic accessibility

FG+HIA+QED+SAs

The molecule contains [FG]. It can be absorbed by human intestinal. It has a high QED score. It has good synthetic accessibility

BTK

The molecule can bind to BTK

FGFR4

The molecule can bind to FGFR4

KPCD3

The molecule can bind to KPCD3

3CL

The molecule can bind to 3CL