Automated and modular protein binder design with BinderFlow

Nayim González-Rodríguez, Carlos Chacón-Sánchez, Oscar Llorca and Rafael Fernández-Leiro

PLOS Computational Biology, 2025, vol. 21, issue 11, 1-14

Abstract: Deep learning has revolutionised de novo protein design, with new models achieving unprecedented success in creating novel proteins with specific functions, including artificial protein binders. However, current workflows remain computationally demanding and challenging to operate without dedicated infrastructure and expertise. To overcome these limitations, we present BinderFlow, an open, structured, and parallelised pipeline that automates end-to-end protein binder design. Its batch-based architecture enables live monitoring of design campaigns, seamless coexistence with other GPU-intensive processes, and minimal user intervention. BinderFlow’s modular design facilitates the integration of new tools, allowing rapid adaptation to emerging methods. We demonstrate its utility by running automated design campaigns that rapidly generate diverse, high-confidence candidates suitable for experimental validation. To complement the pipeline, we developed BFmonitor, a web-based dashboard for real-time campaign monitoring, design evaluation, and hit selection. Together, BinderFlow and BFmonitor make generative protein design more accessible, scalable, and reproducible, streamlining both exploratory and production-level research. The software is freely available at https://github.com/cryoEM-CNIO/BinderFlow under the GNU LGPL v3.0 license.Author summary: The design of artificial proteins that specifically bind protein targets is a promising strategy for developing new therapeutics and research tools. However, current computational pipelines for binder design are complex to operate and rely on large-scale computing resources. Here, we present BinderFlow, a modular and parallelised workflow that simplifies de novo protein binder design. By dividing design campaigns into small, independent batches, BinderFlow allows efficient use of available GPUs, granular control of computational resources, and real-time monitoring. To fully take advantage of this architecture, we built BFmonitor, a web-based interface to visualise campaign metrics, evaluate design quality, and extract promising candidates for experimental validation in real time. We expect both BinderFlow and BFmonitor to make protein design more accessible, enabling researchers from diverse scientific backgrounds to engage directly in the design and refinement of de novo protein binders.

Date: 2025
References: View complete reference list from CitEc
Citations:

Downloads: (external link)
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013747 (text/html)
https://journals.plos.org/ploscompbiol/article/fil ... 13747&type=printable (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:plo:pcbi00:1013747

DOI: 10.1371/journal.pcbi.1013747

Access Statistics for this article

More articles in PLOS Computational Biology from Public Library of Science
Bibliographic data for series maintained by ploscompbiol ().