A recent paper on ChemRxiv https://chemrxiv.org/doi/10.26434/chemrxiv-2025-9c1v6 describes ANNalog a transformer-based sequence-to-sequence generative model trained on pairs of molecules extracted from the same bioactivity assay in a paper as recorded in ChEMBL33.
The idea being this will generate structurally similar analogues involving minor modifications, or bioisosteric replacements typically used in drug discovery. In addition scaffold hopping is also supported.
All the code is on GitHub https://github.com/DVNecromancer/ANNalog
Note the PyPI package does not pin or install PyTorch for you. Please install a PyTorch build that matches your system (CPU/CUDA).
There is also a Google Colab page you can use to try it out.
https://colab.research.google.com/drive/1aJhaBOG7xuYFwMGzfUmbMsLe8T462Ptc#scrollTo=Ss1QOzXjzKSP