AFAaron Lee Feller

Aaron Lee Feller

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Aaron Lee Feller has a total of 31 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Combining Machine Learning and Directed Evolution for Optimization of a Monooxygenase

    by Daniel Gutierrez, Isa Madrigal Harrison, Aaron Feller, Andrew Ellington

    L-3,4-dihydroxyphenylalanine (L-Dopa) is an important pharmaceutical for the treatment of Parkinson's disease and a precursor to numerous catechol-containing compounds. The flavin-dependent monooxygenase HpaBC is a promising biocatalyst for microbial L-Dopa production but exhibits limited native activity toward L-tyrosine. Although structure-based machine learning (ML) models have become increasingly popular for protein engineering, relatively few studies have systematically compared their performance or evaluated their integration into iterative engineering workflows. Here, we benchmarked multiple ML models for their ability to predict activity enhancing mutations in HpaBC. Experimentally validated single mutants were used to seed combinatorial design with EVOLVEpro, generating progressively improved higher-order variants. We next evaluated how expanding the EVOLVEpro training set with directed evolution derived variants influenced combinatorial predictions and finally explored an expanded sequence space by allowing combinations of both machine learning derived and directed evolution derived mutations. This workflow produced HpaBC variants with substantially improved activity. Although incorporating directed evolution data substantially altered EVOLVEpro's predicted mutational trajectories, both training strategies converged on variants with comparable activities, demonstrating that distinct regions of sequence space can yield similarly optimized enzymes. Together, these results provide a systematic comparison of zero-shot ML models and establish an iterative framework for integrating machine learning with directed evolution to accelerate enzyme engineering.

  • Scaling SMILES-Based Chemical Language Models for Therapeutic Peptide Engineering

    by Aaron L. Feller, Maxim Secor, Sebastian Swanson, Claus O. Wilke, Kristine Deibler

    Therapeutic peptides occupy a unique middle ground in drug discovery, offering the high specificity of protein interactions with the chemical diversity of small molecules, yet they currently fall in a computational blind spot. Existing foundation models cannot handle them effectively: protein models are restricted to natural amino acids, while chemical models struggle to process large, polymer-like sequences. This disconnect has forced the field to rely on static chemical descriptors that fail to capture subtle chemical details or on complex multiembedding pipelines that are custom tailored to specific data sets. To bridge this gap, we present PeptideCLM-2, a suite of chemical language models trained on over 100 million molecules to natively represent complex peptide chemistry. This modeling approach expands the available toolkit of machine learning models for therapeutic peptides. Benchmarking results show strong performance versus prior methods for predicting development end points including membrane diffusion, biological function, and half-life.


  • Open Access

    Overestimating zero-shot fitness prediction: Broad benchmarks mask local failures and practical limitations

    by Phillip R. Woolley, Aaron L. Feller, Andrew D. Ellington, Claus O. Wilke

    Deep learning models have emerged as promising tools in protein engineering. In particular, they can be used to predict mutation fitness without the need for task-specific training, a process known as zero-shot prediction. However, the respective strengths and limitations of zero-shot predictions remain poorly understood. Here, we argue that commonly used large-scale benchmarks obscure important failure modes relevant to practical protein engineering, including an inability to pinpoint highly fit mutations or variants driving new-to-nature functions. Beyond these practical failures, we identify a fundamental limitation of zero-shot prediction: a generic fitness score cannot simultaneously optimize for distinct, competing engineering targets, meaning it is inherently disconnected from the phenotype of interest. Moreover, in a systematic comparison of a wide range of available models, we demonstrate that most models show comparable zero-shot performance, irrespective of model architecture and/or input modality (sequence vs. structure). Ultimately, we find that zero-shot predictions serve only as coarse filters separating fit mutations from deleterious ones, failing to reliably identify the mutations that would be most valuable in protein engineering.


  • Open Access

    Validation and analysis of 12,000 AI-driven CAR-T designs in the Bits to Binders competition

    by Clayton W. Kosonocky, Alex M. Abel, Aaron L. Feller, Amanda E. Cifuentes Rieffer, Phillip R. Woolley, Jakub Lála, Daryl R. Barth, Tynan Gardner, Bits To Binders Competitors, Stephen C. Ekker, et al.

    Artificial intelligence (AI) methods for proteins have advanced rapidly, improving structure prediction and design, particularly for de novo binders. However, most evaluations emphasize binding affinity rather than higher-order biological function. We present Bits to Binders , a global competition benchmarking de novo binder design in the context of chimeric antigen receptor (CAR) T cells. Teams from 42 countries submitted 12,000 designs of 80-amino acid binders targeting human CD20 as CAR binding domains. Designs were screened by pooled CAR-T proliferation, identifying 707 designs exhibiting significant CD20-specific enrichment, with team hit rates from 0.6% to 38.4%. Top-performing candidates were validated as individual constructs, measuring CD20-specific proliferation, expansion, cytokine production, and targeted cell lysis. We identified common design methodologies and factors correlated with DNA synthesis, expression, and target-specific T cell activation which nearly double the success rates when applied as a retrospective filter. We release this dataset as an open resource, with practical recommendations to support more effective AI-driven binder design.


  • Open Access

    Scaling SMILES-based chemical language models for therapeutic peptide engineering

    by Aaron L. Feller, Maxim Secor, Sebastian Swanson, Claus O. Wilke, Kristine Deibler

    Therapeutic peptides occupy a unique middle ground in drug discovery, offering the high specificity of protein interactions with the chemical diversity of small molecules, yet they currently fall in a computational blind spot. Existing foundation models cannot handle them effectively: protein models are restricted to natural amino acids, while chemical models struggle to process large, polymer-like sequences. This disconnect has forced the field to rely on static chemical descriptors that fail to capture subtle chemical details or on complex multi-embedding pipelines that are custom tailored to specific datasets. To bridge this gap, we present PeptideCLM-2, a suite of chemical language models trained on over 100 million molecules to natively represent complex peptide chemistry. This modeling approach expands the available toolkit of machine learning models for therapeutic peptides. Benchmarking results show strong performance versus prior methods for predicting development endpoints including membrane diffusion, biological function, and half life.


  • Open Access

    Antibacterial microcins are ubiquitous and functionally diverse across bacterial communities

    by Jennifer K. Parker, Aaron L. Feller, Richard Gu, Simon Sanchez-Paiva, Bethany C. Perez, Angela C. O’dOnnell, Wendi Deng, Rita M. Ousterhout, Sun-Young Kim, Claus O. Wilke, et al.

    Bacteria produce antibacterials that drive competition and regulate community composition. While diverse examples have been found, few families of antibacterial agents appear to be widespread across phylogenetically divergent bacteria. Here, we show that what appeared to be a limited, niche class of Gram-negative bacteriocins, called class II microcins, is in fact a highly abundant, sequence- and function-diverse class of secreted bacteriocins. Based on systematic investigations in the Enterobacteriaceae and gut microbiomes, we demonstrate that class II microcins encompass diverse sequence space, bacterial strains of origin, spectra of activity, and mechanisms of action. Importantly, we show microcins discovered here are active against pathogenic E. coli during mouse gut colonization, supporting important roles for these unrecognized antibacterials in vivo. Our study reveals the overlooked abundance and diversity of microcins found dispersed throughout Bacteria and opens opportunities to uncover and exploit mechanisms of competition to modulate microbial communities.


  • Peptide-Aware Chemical Language Model Successfully Predicts Membrane Diffusion of Cyclic Peptides

    by Aaron L. Feller, Claus O. Wilke

    Language modeling applied to biological data has significantly advanced the prediction of membrane penetration for small-molecule drugs and natural peptides. However, accurately predicting membrane diffusion for peptides with pharmacologically relevant modifications remains a substantial challenge. Here, we introduce PeptideCLM, a peptide-focused chemical language model capable of encoding peptides with chemical modifications, unnatural or noncanonical amino acids, and cyclizations. We assess this model by predicting membrane diffusion of cyclic peptides, demonstrating greater predictive power than existing chemical language models. Our model is versatile and can be extended beyond membrane diffusion predictions to other target values. Its advantages include the ability to model macromolecules using chemical string notation, a largely unexplored domain, and a simple, flexible architecture that allows for adaptation to any peptide or other macromolecule data set.


  • Open Access

    Peptide-aware chemical language model successfully predicts membrane diffusion of cyclic peptides

    by Aaron Lee Feller, Claus O. Wilke

    Language modeling applied to biological data has significantly advanced the prediction of membrane penetration for small molecule drugs and natural peptides. However, accurately predicting membrane diffusion for peptides with pharmacologically relevant modifications remains a substantial challenge. Here, we introduce PeptideCLM, a peptide-focused chemical language model capable of encoding peptides with chemical modifications, unnatural or non-canonical amino acids, and cyclizations. We assess this model by predicting membrane diffusion of cyclic peptides, demonstrating greater predictive power than existing chemical language models. Our model is versatile and can be extended beyond membrane diffusion predictions to other target values. Its advantages include the ability to model macromolecules using chemical string notation, a largely unexplored domain, and a simple, flexible architecture that allows for adaptation to any peptide or other macromolecule dataset.


  • Open Access

    Evidence for Widespread Class II Microcins in Enterobacterales Genomes

    by T. Jeffrey Cole, Jennifer K. Parker, Aaron L. Feller, Claus O. Wilke, Bryan W. Davies

    Class II microcins are small bacteriocins produced by strains of Gram-negative bacteria in the Enterobacteriaceae . They are generally understood to play a role in interbacterial competition, although direct evidence of this is limited, and they could prove informative in developing new peptide antibiotics.