Table 6 Performance evaluation of different models after fine tuning in ‘-tinib’ data set for conditioned generation task@.

From: Screening of multi deep learning-based de novo molecular generation models and their application for specific target molecular generation

Model

Validity

Uniqueness

Novelty

Similarity ratio

T5MolGe

0.884 ± 0.001

0.614 ± 0.005

0.965 ± 0.012

0.963 ± 0.013

GPT-RoPE

0.744 ± 0.011

0.820 ± 0.008

0.992 ± 0.033

0.910 ± 0.027

  1. @Mean standard deviation (mean \(\pm\) SD) of Validity, Uniqueness, Novelty, FCD, and KL for five random seeds experiments.
Back to article page