Cited by Design of multimodal antibiotics against intracellular infections using deep learning

Cell-autonomous innate immunity by proteasome-derived defence peptides DOI

Karin Goldberg, Arseniy Lobov, Paola Antonello

et al.

Nature, Journal Year: 2025, Volume and Issue: unknown

Published: March 5, 2025

For decades, antigen presentation on major histocompatibility complex class I for T cell-mediated immunity has been considered the primary function of proteasome-derived peptides1,2. However, whether products proteasomal degradation play additional parts in mounting immune responses remains unknown. Antimicrobial peptides serve as a first line defence against invading pathogens before adaptive system responds. Although protective antimicrobial across numerous tissues is well established, cellular mechanisms underlying their generation are not fully understood. Here we uncover role proteasomes constitutive and bacterial-induced that impede bacterial growth both vitro vivo by disrupting membranes. In silico prediction proteome-wide cleavage identified hundreds thousands potential with cationic properties may be generated en route to act defence. Furthermore, infection induces changes proteasome composition function, including PSME3 recruitment increased tryptic-like cleavage, enhancing activity. Beyond providing mechanistic insights into cell-autonomous innate immunity, our study suggests proteasome-cleaved have previously overlooked functions downstream degradation. From translational standpoint, identifying could provide an untapped source natural antibiotics biotechnological applications therapeutic interventions infectious diseases immunocompromised conditions.

Language: Английский

Citations

Generative latent diffusion language modeling yields anti-infective synthetic peptides DOI

Marcelo D. T. Torres, Tianlai Chen, Fangping Wan

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 1, 2025

Abstract Generative artificial intelligence (AI) offers a powerful avenue for peptide design, yet this process remains challenging due to the vast sequence space, complex structure–activity relationships, and need balance antimicrobial potency with low toxicity. Traditional approaches often rely on trial-and-error screening fail efficiently navigate immense diversity of potential sequences. Here, we introduce AMP-Diffusion, novel latent diffusion model fine-tuned (AMP) sequences using embeddings from protein language models. By systematically exploring AMP-Diffusion enables rapid discovery promising antibiotic candidates. We generated 50,000 candidate sequences, which were subsequently filtered ranked our APEX predictor model. From these, 46 top candidates synthesized experimentally validated. The resulting peptides demonstrated broad-spectrum antibacterial activity, targeting clinically relevant pathogens—including multidrug-resistant strains—while exhibiting cytotoxicity in human cell assays. Mechanistic studies revealed bacterial killing via membrane permeabilization depolarization, showed favorable physicochemical profiles. In preclinical mouse models infection, lead effectively reduced burdens, displaying efficacy comparable polymyxin B levofloxacin, no detectable adverse effects. This study highlights as robust generative platform designing antibiotics bioactive peptides, offering strategy address escalating challenge resistance.

Language: Английский

Citations

Frog-derived synthetic peptides display anti-infective activity against Gram-negative pathogens DOI

Lucía Ageitos, Andreia Boaro, Angela Cesaro

et al.

Trends in biotechnology, Journal Year: 2025, Volume and Issue: unknown

Published: March 1, 2025

Novel antibiotics are urgently needed since bacteria becoming increasingly resistant to existing antimicrobial drugs. Furthermore, available broad spectrum, often causing off-target effects on host cells and the beneficial microbiome. To overcome these limitations, we used structure-guided design generate synthetic peptides derived from Andersonin-D1, an peptide (AMP) produced by odorous frog Odorrana andersonii. We found that both hydrophobicity net charge were critical for its bioactivity, enabling of novel, optimized peptides. These selectively targeted Gram-negative pathogens in single cultures complex microbial consortia, showed no human or gut microbes, did not select bacterial resistance. Notably, they also exhibited vivo activity two preclinical murine models. Overall, present target pathogenic infections offer promising antibiotic candidates.

Language: Английский

Citations

Discovery of antibiotics in the archaeome using deep learning DOI

Marcelo D. T. Torres, Fangping Wan, César de la Fuente‐Núñez

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 16, 2024

Antimicrobial resistance (AMR) is one of the greatest threats facing humanity, making need for new antibiotics more critical than ever. While most have traditionally been derived from bacteria and fungi, archaea-a distinct underexplored domain life-offer a largely untapped reservoir antibiotic discovery. In this study, we leveraged deep learning to systematically explore archaeome, uncovering promising candidates combating AMR. By mining 233 archaeal proteomes, identified 12,623 molecules with potential antimicrobial activity. These newly discovered peptide compounds, termed archaeasins, exhibit unique compositional features that differentiate them traditional peptides, including amino acid profile. We synthesized 80 93% which demonstrated activity

Language: Английский

Citations

A generative artificial intelligence approach for antibiotic optimization DOI

Marcelo D. T. Torres,

Yimeng Zeng,

Fangping Wan

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 27, 2024

Abstract Antimicrobial resistance (AMR) poses a critical global health threat, underscoring the urgent need for innovative antibiotic discovery strategies. While recent advances in peptide design have yielded numerous antimicrobial agents, optimizing these molecules experimentally remains challenging due to unpredictable and resource-intensive trial-and-error approaches. Here, we introduce APEX Generative Optimization (APEX GO ), generative artificial intelligence (AI) framework that integrates transformer-based variational autoencoder with Bayesian optimization optimize peptides. Unlike traditional supervised learning approaches screen fixed databases of existing molecules, generates entirely novel sequences through arbitrary modifications template peptides, representing paradigm shift discovery. Our introduces new diversity constraints maintain similarity specific templates while enabling sequence innovation. This work represents first vitro vivo experimental validation any setting. Using ten de-extinct peptides as templates, generated optimized derivatives enhanced properties. We synthesized 100 conducted comprehensive characterizations, including assessments activity, mechanism action, secondary structure, cytotoxicity. Notably, achieved an outstanding 85% ground-truth hit rate 72% success enhancing activity against clinically relevant Gram-negative pathogens, outperforming previously reported methods optimization. In preclinical mouse models Acinetobacter baumannii infection, several AI-optimized molecules—most notably mammuthusin-3 mylodonin-2—exhibited potent anti-infective comparable or exceeding polymyxin B, widely used last-resort antibiotic. These findings highlight potential AI approach optimization, offering powerful tool accelerate address escalating challenge AMR.

Language: Английский

Citations

Venomics AI: a computational exploration of global venoms for antibiotic discovery DOI

Changge Guan, Marcelo D. T. Torres,

Sufen Li

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 20, 2024

The relentless emergence of antibiotic-resistant pathogens, particularly Gram-negative bacteria, highlights the urgent need for novel therapeutic interventions. Drug-resistant infections account approximately 5 million deaths annually, yet antibiotic development pipeline has largely stagnated. Venoms, representing a remarkably diverse reservoir bioactive molecules, remain an underexploited source potential antimicrobials. Venom-derived peptides, in particular, hold promise discovery due to their evolutionary diversity and unique pharmacological profiles. In this study, we mined comprehensive global venomics datasets identify new antimicrobial candidates. Using machine learning, explored 16,123 venom proteins, generating 40,626,260 venom-encrypted peptides (VEPs). APEX, deep learning model combining peptide-sequence encoder with neural networks activity prediction, identified 386 VEPs structurally functionally distinct from known peptides. Our analyses showed that these possess high net charge elevated hydrophobicity, characteristics conducive bacterial membrane disruption. Structural studies revealed considerable conformational flexibility, many transitioning α-helical conformations membrane-mimicking environments, indicative potential. Of 58 selected experimental validation, 53 displayed potent activity. Mechanistic assays indicated primarily exert effects through depolarization, mirroring AMP-like mechanisms. vivo using mouse Acinetobacter baumannii infection demonstrated lead significantly reduced burdens without notable toxicity. This study value venoms as resource antibiotics. By integrating computational approaches venom-derived emerge promising candidates combat challenge resistance.

Language: Английский

Citations

Design of multimodal antibiotics against intracellular infections using deep learning DOI

Angela Cesaro, Fangping Wan, Marcelo D. T. Torres

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 21, 2024

Abstract The rise of antimicrobial resistance has rendered many treatments ineffective, posing serious public health challenges. Intracellular infections are particularly difficult to treat since conventional antibiotics fail neutralize pathogens hidden within human cells. However, designing molecules that penetrate cells while retaining activity historically been a major challenge. Here, we introduce APEX DUO , multimodal artificial intelligence (AI) model for generating peptides with both cell-penetrating and properties. From library 50 million AI-generated compounds, selected characterized several candidates. Our lead, Turingcin, penetrated mammalian eradicated intracellular Staphylococcus aureus . In mouse models skin abscess peritonitis, Turingcin reduced bacterial loads by up two orders magnitude. sum, generated antibiotics, opening new avenues molecular design.

Language: Английский

Citations