Aug 2026· Research· Vol 9· 0 citations· 43 references
Medicine
TL;DR
Mirror-Peptidizer is reported, an end-to-end in silico mirror-image screening workflow that generates D-peptide binders without requiring chemical synthesis of D-protein targets.
Abstract
D-peptides are attractive therapeutic modalities because they are generally more resistant to proteolysis than their L-counterparts, yet systematic design of target-binding D-peptides remains nontrivial. Here, we report Mirror-Peptidizer, an end-to-end in silico mirror-image screening workflow that generates D-peptide binders without requiring chemical synthesis of D-protein targets. The workflow mirrors an L-protein structure to a virtual D-protein, designs L-peptide backbones in the presence of the mirrored target using a diffusion-based backbone generator, selects sequences with a neural sequence design model, and explores local sequence neighborhoods via Bayesian multi-objective optimization balancing sequence-backbone compatibility and a solubility heuristic. Mirroring the resulting complex yields the corresponding D-peptide predicted to bind the native L-target. Using MDM2, PD-L1, and interleukin-23 receptor (IL-23R) as test cases, we identified D-peptides spanning α-helical, β-rich, and mixed conformations with affinity from 11.9 nM to sub-μM. For the MDM2 system, the 1H-15N HSQC (heteronuclear single quantum coherence) perturbations and protein mutagenesis support the designed interface, and cell-penetrating conjugates show p53-dependent cancer growth inhibition. Similarly, PD-L1-targeting D-peptides potently inhibited PD-1/PD-L1 interactions in competitive binding assays in vitro, and IL-23R-targeting D-peptides inhibited IL-2/IL-12/IL-23-induced interferon-γ production in human peripheral blood mononuclear cells. Mirror-Peptidizer is provided as an open-source implementation to facilitate rapid generation of experimentally testable D-peptide starting points.
D-peptides combine protease resistance with high target specificity, but computational design of D-peptide binders remains immature. Mirror-Peptidizer introduced an in silico mirror-image screening pipeline using target reflection, backbone generation, and ProteinMPNN sequence design, but its raw ProteinMPNN negative l...
HighMorph is presented, an interaction-guided framework that combines protein–protein interaction information with artificial intelligence for rational cyclic peptide design and provides insights for developing therapeutics targeting challenging protein interfaces.
M. Lan, Cheng-Yun Zhang, Wentong Wang et al.· Journal of Medicinal Chemist...· 0 citations
Intrinsically disordered proteins, topologically complex protein surfaces, and biomolecular condensates regulate essential cellular processes but remain difficult to drug with conventional small molecules. Here, we present PepSpace, a physics-driven directed-evolution platform coupling residue-resolution coarse-grained...
J. Oller-Iscar, A. Tejedor, Alejandro Feito et al.· bioRxiv· 0 citations
The REceptor-Aware Peptide Sequence Designer (REAPS), a geometric graph neural network that reframes sequence design for peptide binders by treating the receptor as a fully observable all-atom context, is introduced.
Yong-Kang Qiu, Jing-Xin Qiao, Leng-Jing Zhu et al.· Journal of Chemical Informat...· 0 citations
Peptide lasso-grafting strategies enable the transfer of established pharmacophores into compact protein scaffolds but often require optimization to preserve target binding and biological activity. Here, we report Random in-vitro Protein single-Loop Engineering (RiPLE), an mRNA-display strategy for the direct discovery...
Kilian Colas, T. Ko, G. Philippe et al.· bioRxiv· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 20, 2026
Senior Research Scientist Jim Glass discusses how his MIT-IBM Computing Research Lab collaboration connects perception and language, and addresses the shortcoming of current models.
MIT News · Artificial Intelligence· news.mit.eduNov 20, 2026
Enforcing physics-based constraints prevents AI models from generating unreliable predictions for complex nonlinear systems in weather, climate, and turbulent flows.
An MIT and Dana-Farber Institute-led team used large language models to examine nearly 4,000 studies and hundreds of code repositories, revealing gaps in the information needed to scrutinize and reuse published research.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.