Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated 52 min ago
Peer-reviewedCC BY 4.0Bioinformatics

Multi-objective-conditioned-return-blocks decision transformer: a novel approach for multi-objective de novo drug design

  • Eyal Mazuz1
  • Or Gindes1
  • Lior Rokach1
  • Bracha Shapira1
  1. 1Faculty of Computer and Information Sciences, Ben-Gurion University of the Negev , Be'er Sheva 8410501,
Open at sourcePDFDOICite

Abstract

Motivation: Designing novel molecules that simultaneously satisfy several, often conflicting, properties is a long-standing challenge in drug design. While recent deep learning advances optimize single targets effectively, navigating the tradeoffs required for multiple objectives remains a difficult problem.

Results: We introduce the Multi-Objective-Conditioned-Return-Blocks Decision Transformer (M-GoRT), a novel architecture for multi-target de-novo drug design incorporating return-to-go blocks conditioned on goal embeddings. Trained without exploration, reward weighting or online fine-tuning, M-GoRT efficiently optimizes for multiple targets. M-GoRT achieves competitive multi-objective success rates against state-of-the-art baselines while maintaining superior structural diversity and novelty, even in high-dimensional tasks incorporating biological targets (DRD2, JNK3). In these high-dimensional tasks, M-GoRT maintains superior structural diversity and novelty establishing it as a practical tool for multi-property molecular generation.

Availability and implementation: M-GoRT and the data used are freely available as an open-source software under the MIT license at GitHub https://github.com/eyalmazuz/MolGen/tree/GCRB-DT.

Related

  1. Motivationally objective versus subjective decision-making: Neural correlates, behavior, and self-report
    P. Modak, J. W. Brown
    bioRxivPreprintCC BY-NDFull textNeurosciencebioRxiv
  2. A broadly active flavivirus NS4B targeting compound that blocks NS4B-NS3 binding and replication organelle formation
    Jessica L Smith, Theresa H. Nguyen, Sixue Zhang +3
    bioRxivPreprintAll rights reservedFull textMicrobiologybioRxivPMC
  3. Targeted RNA degradation of HIV-1 TAR element at 5'UTR by CRISPR-Cas13X blocks HIV-1 lytic reactivation
    T. T. Lepcha, D. Zhou, G. N. Fiches +7
    bioRxivPreprintAll rights reservedFull textMicrobiologybioRxiv
  4. Bundling Alleles Within Haplotype Blocks Improves QTL Detection Under Allelic Heterogeneity
    S. Oome, M. Ghanbari
    bioRxivPreprintCC BYFull textGenetics & genomicsbioRxiv
  5. Long-term decline in egg size of Japanese hatchery chum salmon links to declines in return rate and abundance
    S. Kitada
    bioRxivPreprintAll rights reservedFull text1 citedEvolutionary biologybioRxiv

Cite

BibTeXRIS
BibTeX
@article{mazuz2026multiobjectiveconditionedreturnblocks,
  title = {{Multi-objective-conditioned-return-blocks decision transformer: a novel approach for multi-objective de novo drug design}},
  author = {Eyal Mazuz and Or Gindes and Lior Rokach and Bracha Shapira},
  journal = {Bioinformatics},
  year = {2026},
  month = oct,
  doi = {10.1093/bioinformatics/btag731},
  url = {https://academic.oup.com/bioinformatics/article/doi/10.1093/bioinformatics/btag731/8860138},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Multi-objective-conditioned-return-blocks decision transformer: a novel approach for multi-objective de novo drug design
AU  - Eyal Mazuz
AU  - Or Gindes
AU  - Lior Rokach
AU  - Bracha Shapira
PY  - 2026
DA  - 2026/10/01
JO  - Bioinformatics
DO  - 10.1093/bioinformatics/btag731
UR  - https://academic.oup.com/bioinformatics/article/doi/10.1093/bioinformatics/btag731/8860138
AB  - Motivation: Designing novel molecules that simultaneously satisfy several, often conflicting, properties is a long-standing challenge in drug design. While recent deep learning advances optimize single targets effectively, navigating the tradeoffs required for multiple objectives remains a difficult problem.  Results: We introduce the Multi-Objective-Conditioned-Return-Blocks Decision Transformer (M-GoRT), a novel architecture for multi-target de-novo drug design incorporating return-to-go blocks conditioned on goal embeddings. Trained without exploration, reward weighting or online fine-tuning, M-GoRT efficiently optimizes for multiple targets. M-GoRT achieves competitive multi-objective success rates against state-of-the-art baselines while maintaining superior structural diversity and novelty, even in high-dimensional tasks incorporating biological targets (DRD2, JNK3). In these high-dimensional tasks, M-GoRT maintains superior structural diversity and novelty establishing it as a practical tool for multi-property molecular generation.  Availability and implementation: M-GoRT and the data used are freely available as an open-source software under the MIT license at GitHub https://github.com/eyalmazuz/MolGen/tree/GCRB-DT.
SN  - 1367-4803
SN  - 1367-4811
ER  -