Biology subjects

Romero, P. A.

Publications and source records attributed to Romero, P. A..

At least 19 recordsLinked to original sources

bioRxiv2026

Glioblastoma model mouse rescue from mortality by brain-targeted catalytically-enhanced angiotensin-converting enzyme 2 mutant-antibody fusion protein

We have discovered that an engineered angiotensin-converting enzyme 2 (ACE2)-brain-targeted antibody fusion protein, referred to as ACE2dm-s1, improves survival of CT-2A xenograft glioblastoma (GBM) model mice when administered in combination with a small molecule STING agonist cancer therapeutic. Particularly striking about these survival results is that prior efforts that used losartan, a small molecule angiotensin receptor antagonist that acts on the same Renin-Angiotensin System (RAS) axis that is affected by the action of ACE2dm-s1, showed that this small molecule provided no survival benefit when used as an element of combination therapy in the same CT-2A GBM model used in our experiments. The improved CT-2A mouse survival, which was not observed for ACE2dm-s1 or STING agonist monotherapy, combines with our flow cytometry study results that revealed an ACE2dm-s1-mediated shift of brain immune cell populations toward an anti-tumoral composition to establish ACE2dm-s1's utility as a tumor microenvironment (TME) modulator that could serve as an adjuvant in increasing the efficacy of wide range of cancer treatment modalities including chemotherapeutic agents, immune checkpoint inhibitors, and CAR-T cells.

cancer biology↗

bioRxiv2026

Experimental reasoning in large language models

Scientific reasoning requires interaction with the physical world: designing experiments, interpreting outcomes, and deciding what experiment should come next. Here we show that general-purpose large language models (LLMs) already possess substantial capabilities for experimental reasoning. We connected LLMs to a liquid-handling robot and camera through PEARL, which provides perception, action, and persistent state. Given objectives rather than prescribed workflows, the system designed and successfully executed experiments, interpreted observations, and revised its approach as evidence accumulated. When information was insufficient, PEARL autonomously designed and performed intermediate experiments to acquire it. In quantitative enzyme characterization, it judged whether measurements were sufficient, new data were required, or conditions needed to change. Across eight scenarios, these behaviors were reliable with limited agent scaffolding and no supplied scenario-specific recovery recipes. More broadly, general-purpose LLMs may provide a reasoning layer between scientific objectives and physical experimentation, shifting laboratory automation from executing predefined procedures toward AI systems that participate directly in the experimental process.

bioengineering↗

bioRxiv2026

Megascale recombination generates millions of catalytically competent PETases

Engineering proteins at the scale of modern high-throughput screening technologies remains challenging because most experimental libraries sample only limited regions of sequence space while maintaining relatively low levels of functional diversity. Although recombination-based strategies can introduce substantially larger sequence changes than conventional mutagenesis approaches, it is unclear whether highly recombined enzymes can retain catalytic competence when library sizes approach millions to billions of variants. To investigate this question, we constructed a recombination library by combining ten sequence blocks from ten bacterial PETases, generating a theoretical diversity of 10^8 chimeric enzymes. Screening of yeast-displayed libraries using a fluorophosphonate probe for catalytic-serine reactivity identified substantial populations of catalytically competent variants despite extensive sequence recombination and dozens of amino acid substitutions per chimera. Approximately 2% of sampled variants possessed active-site serine reactivity, corresponding to an estimated 10 catalytically competent enzymes within the full theoretical library. Deep sequencing showed that functional populations remained broadly distributed across sequence space following selection and contained multiple enriched recombination patterns spanning the PETase structure. Secondary lysate screening and enzyme-normalized activity measurements further identified multiple purified PETase chimeras with soluble esterase activity comparable to or greater than the wild type Ideonella sakaiensis PETase on a chromogenic surrogate substrate. Together, these results demonstrate that large-scale recombination can generate millions of diverse and catalytically competent enzymes while remaining compatible with modern high-throughput screening workflows. More broadly, this work establishes recombination-based diversity generation as a scalable strategy for protein engineering and large-scale sequence-function data generation.

bioengineering↗

bioRxiv2026

Integrating complementary biological information for multi-objective enzyme engineering

Enzyme catalysts are increasingly used for sustainable pharmaceutical manufacturing, but engineering industrial biocatalysts remains challenging as multiple catalytic and developability properties must be optimized simultaneously from limited experimental data. Here, we develop a machine learning-guided multi-objective design framework that integrates sparse functional measurements with complementary evolutionary and structural information to engineer the ketoreductase Gre2. The resulting designs achieved simultaneous improvements in catalytic performance, protein yield, and thermal stability, with selected designs retaining improved performance under process-relevant conditions. More broadly, our results demonstrate that integrating complementary biological information enables efficient multi-objective enzyme engineering from sparse experimental data, providing a general strategy for accelerating industrial biocatalyst development.

bioengineering↗

bioRxiv2026

Learning protein function through autonomous experimental interaction

Biological AI learns primarily from existing observations, but many questions cannot be answered from available data alone. Here we show that AI can instead acquire knowledge by acting directly on biological systems and learning from the consequences. We developed a closed-loop framework in which autonomous agents design protein variants, construct and characterize them in a robotic laboratory, learn from the resulting experimental feedback, and decide what experiments to perform next. We then allowed the system to operate continuously and without human intervention for approximately one month, during which multiple agents independently explored protein sequence space while learning from shared experimental experience. Applied to glycoside hydrolases, the agents discovered enzymes with substantially altered substrate specificity toward non-native sugars and progressively learned the structure of the underlying sequence-function landscape. The resulting experimental experience also revealed determinants of substrate specificity and protein expression that were not specified as learning objectives. These results demonstrate that AI can autonomously interact with biology over extended periods to acquire knowledge through experience, establishing a framework for biological discovery driven by continuous experimental interaction.

bioengineering↗

bioRxiv2026

Machine learning-guided olivetolic acid cyclase engineering enables tailored cannabinoid biosynthesis in yeast

Cannabinoids comprise a diverse class of bioactive natural products with important therapeutic potential, but efficient microbial production remains limited by pathway bottlenecks and challenges in engineering key biosynthetic enzymes. Here, we develop a machine learning-guided approach to engineer olivetolic acid cyclase (OAC), a critical control point in cannabinoid biosynthesis that governs both pathway flux and product selectivity. We first generated sequence-function data from 152 CsOAC variants spanning homolog screening, recombination, and mutagenesis libraries. Using these measurements, we trained multi-task models to predict pathway-level production of olivetolic acid (OA), divarinic acid (DVA), and competing byproducts, together with a variational autoencoder that captured evolutionary constraints across the broader enzyme family. Across three rounds of iterative design and testing, this approach identified CsOAC variants that substantially increased production and selectivity of both OA and DVA. When introduced into engineered Yarrowia lipolytica strains, these variants enabled production of tetrahydrocannabinolic acid (THCA) and the minor cannabinoid tetrahydrocannabivarinic acid (THCVA) at titers exceeding previous yeast systems. Analysis of top-performing variants revealed mutations influencing substrate selectivity and catalytic performance, providing insight into the determinants of CsOAC function. More broadly, this work demonstrates how machine learning-guided enzyme engineering can improve pathway performance and expand access to major and minor cannabinoids through microbial biosynthesis.

bioengineering↗

bioRxiv2026

Explicit representation of germline and non-germline residues improves antibody language modeling

Antibodies originate from germline templates and are diversified by somatic hypermutation, producing sequences in which conserved germline residues scaffold structure while rare non-germline (NGL) substitutions refine antigen binding. Current antibody language models (ALMs) treat all residues equivalently and inherit a germline bias that systematically down-weights functionally critical NGL mutations as statistical noise. We introduce PRISM, a germline-aware ALM that explicitly represents germline and nongermline residues as distinct token types over a factorized 53-token vocabulary. PRISM achieves state-of-the-art pseudo-perplexity in hypervariable CDRs and is uniquely positively correlated with experimental binding affinity across three deep mutational scanning landscapes on which all compared ALMs anti-correlate. The dual-vocabulary further enables property-specific controllable generation previously unattainable with entangled ALMs. NGL-directed sampling improves physics-based binding scores while GL-directed sampling preserves stability and solubility. These results establish disentangled germline/non-germline representation as a substantive advance in antibody language modeling.

immunology↗

bioRxiv2026

Benchmarking and behavioral characterization of LLM agents for protein design

Large language models (LLMs) are increasingly deployed as agents for scientific discovery, but standardized frameworks for evaluating their performance and behavior in scientific workflows are lacking. Protein design provides a demanding test case because modern workflows combine stochastic generative models, structure prediction systems, and physics-based evaluation tools that require extensive candidate exploration and filtering. Here we introduce BioDesignBench, a benchmark of 76 expert-curated protein design tasks spanning antibodies, enzymes, fluorescent proteins, binders, and scaffolds, together with human and non-LLM baselines and behavioral metrics derived from tool-use traces. We evaluate four frontier LLM agents across diverse protein design workflows and find that the strongest agents surpass deterministic hardcoded pipelines but consistently underperform expert practice. Decomposing agent behavior along orthogonal axes of tool coverage and evaluation depth, we localize the gap to evaluation rather than tool selection: agents generally pick appropriate tools but score each candidate against only a narrow set of metrics, rarely compare alternatives, and terminate exploration prematurely. Guided workflows improve tool coverage but not evaluation depth. A compute-matched forced-depth intervention that requires agents to evaluate every candidate across multiple complementary metric categories (structure, interface, physics, and learned affinity) substantially improves performance and rules out generic scaffolding effects, demonstrating that the gap is behavioral rather than a fundamental capability constraint. All evaluations are performed in silico. We release BioDesignBench, open-source reference agents, and a public leaderboard as a community resource for evaluating and improving AI agents for protein engineering.

bioengineering↗

bioRxiv2026

AlphaFast: High-throughput AlphaFold 3 via GPU-accelerated MSA construction

AlphaFold 3 (AF3) enables accurate biomolecular modeling but is limited by slow, CPU-bound multiple sequence alignment (MSA) generation. We introduce AlphaFast, a drop-in framework that integrates GPU-accelerated MMseqs2 sequence search to remove this bottleneck. AlphaFast achieves a 68.5x speedup in MSA construction and a 22.8x reduction in end-to-end runtime on a single GPU, and delivers predictions in 8 seconds per input on four GPUs while maintaining indistinguishable structural accuracy. A serverless deployment enables structure prediction for as little as $0.035 per input. Code is available at https://github.com/RomeroLab/alphafast

biochemistry↗

bioRxiv2025

Functional alignment of protein language models via reinforcement learning

Protein language models (pLMs) enable generative design of novel protein sequences but remain fundamentally misaligned with protein engineering goals, as they lack explicit understanding of function and often fail to improve properties beyond those found in nature. We introduce Reinforcement Learning from eXperimental Feedback (RLXF), a general framework that aligns protein language models with experimentally measured functional objectives, drawing inspiration from the methods used to align large language models like ChatGPT. Applied across five diverse protein families, RLXF improves generation of high-functioning variants beyond pre-trained baselines. We demonstrate this with CreiLOV, an oxygen-independent fluorescent protein, where RLXF-aligned models generate sequences with significantly enhanced fluorescence, including the most fluorescent CreiLOV variants reported to date. Our results indicate that RLXF-aligned models effectively integrate the evolutionary knowledge encoded in pre-trained pLMs with experimental observations, improving the success rate of generated sequences and enabling the discovery of synergistic mutation combinations that are difficult to identify through zero-shot or evolutionary approaches. RLXF provides a scalable and accessible approach to steer generative models toward desired biochemical properties, enabling function-driven protein design beyond the limits of natural evolution.

bioengineering↗

bioRxiv2025

Scalable and cost-efficient custom gene library assembly from oligopools

Advances in metagenomics, deep learning, and generative protein design have enabled broad in silico exploration of sequence space, but experimental characterization is still constrained by the cost and scalability of DNA synthesis. Here, we present OMEGA (Oligo-based Multiplexed Efficient Gene Assembly), a low-cost, accessible method for assembling hundreds to thousands of full-length genes in parallel using standard laboratory techniques. OMEGA computationally fragments target genes into short, high-fidelity Golden Gate-compatible oligonucleotides that can be ordered as a pooled library and assembled across multiplexed subpools. We systematically optimized the number of fragments per gene and orthogonal ligation sites per reaction and determine that OMEGA can assemble up to 2.6 kb constructs using as many as 70 Golden Gate sites. To validate the approach, we assembled and functionally screened a library of 810 natural and synthetic GFP variants, recovering 94-97% of target sequences with high uniformity. OMEGA enables precision library construction at scale, with per-gene costs as low as $1.50, and offers a broadly applicable solution for bridging computational protein design with high-throughput experimental validation. We have developed OMEGA as an open-source software package and an easy-to-use Colab notebook available at https://github.com/RomeroLab/omega.

synthetic biology↗

bioRxiv2024

Droplet microfluidic screening to engineer angiotensin-converting enzyme 2 (ACE2) catalytic activity

BackgroundAngiotensin-Converting Enzyme 2 (ACE2) is a crucial peptidase in human peptide hormone signaling, catalyzing the conversion of Angiotensin-II to Angiotensin-(1-7), which activates the Mas receptor and elicits vasodilation, increased blood flow, reduced inflammation, and decreased pathological tissue remodeling. This study leverages protein engineering to enhance ACE2s therapeutic potential for treating conditions such as respiratory viral infections, acute respiratory distress syndrome, and diabetes. Surrogate substrates used in traditional high-throughput screening methods for peptidases often fail to accurately mimic native substrates, leading to less effective enzyme variants. Here, we developed an ultra-high-throughput droplet microfluidic platform to screen peptidases on native peptide substrates. Our assay detects substrate cleavage via free amino acid release, providing a precise measurement of biologically relevant peptidase activity. ResultsUsing this new platform, we screened a large library of ACE2 variants, identifying position 187 as a hotspot for enhancing enzyme activity. Further focused screening revealed the K187T variant, which exhibited a fourfold increase in catalytic efficiency (kcat/KM) over wild-type ACE2. ConclusionsThis work demonstrates the potential of droplet microfluidics for therapeutic peptidase engineering, offering a robust and accessible method to optimize enzyme properties for clinical applications.

bioengineering↗

bioRxiv2024

Engineered lysins to modulate human gut microbiome communities

The ability to precisely control the human gut microbiomes composition, metabolic networks, and host interactions would revolutionize medicine and how we treat human disease. Microbiome remodeling approaches such as probiotics, prebiotics, antibiotics, and fecal microbiota transplantation are global perturbations with consequences that are difficult to predict. Lysins are enzymes produced by bacteriophages and bacteria that have exceptional specificity to lyse bacterial cells and could be applied for targeted microbiome remodeling. In this work we demonstrate the ability of lysins to target diverse human gut commensal bacteria, how domain recombination can give rise to chimeric lysins with altered species preferences, and how engineered lysins can directly remodel the structure and function of synthetic human gut microbiomes. Lysins represent a general platform that can be rapidly tailored through protein engineering to control complex microbial ecosystems.

synthetic biology↗

bioRxiv2024

Biophysics-based protein language models for protein engineering

Protein language models trained on evolutionary data have emerged as powerful tools for predictive problems involving protein sequence, structure, and function. However, these models overlook decades of research into biophysical factors governing protein function. We propose Mutational Effect Transfer Learning (METL), a protein language model framework that unites advanced machine learning and biophysical modeling. Using the METL framework, we pretrain transformer-based neural networks on biophysical simulation data to capture fundamental relationships between protein sequence, structure, and energetics. We finetune METL on experimental sequence-function data to harness these biophysical signals and apply them when predicting protein properties like thermostability, catalytic activity, and fluorescence. METL excels in challenging protein engineering tasks like generalizing from small training sets and position extrapolation, although existing methods that train on evolutionary signals remain powerful for many types of experimental assays. We demonstrate METLs ability to design functional green fluorescent protein variants when trained on only 64 examples, showcasing the potential of biophysics-based protein language models for protein engineering.

bioinformatics↗

bioRxiv2023

Neural network extrapolation to distant regions of the protein fitness landscape

Machine learning (ML) has transformed protein engineering by constructing models of the underlying sequence-function landscape to accelerate the discovery of new biomolecules. ML-guided protein design requires models, trained on local sequence-function information, to accurately predict distant fitness peaks. In this work, we evaluate neural networks capacity to extrapolate beyond their training data. We perform model-guided design using a panel of neural network architectures trained on protein G (GB1)-Immunoglobulin G (IgG) binding data and experimentally test thousands of GB1 designs to systematically evaluate the models extrapolation. We find each model architecture infers markedly different landscapes from the same data, which give rise to unique design preferences. We find simpler models excel in local extrapolation to design high fitness proteins, while more sophisticated convolutional models can venture deep into sequence space to design proteins that fold but are no longer functional. Our findings highlight how each architectures inductive biases prime them to learn different aspects of the protein fitness landscape.

synthetic biology↗

bioRxiv2023

Self-driving laboratories to autonomously navigate the protein fitness landscape

Protein engineering has nearly limitless applications across chemistry, energy, and medicine, but creating new proteins with improved or novel functions remains slow, labor-intensive, and inefficient. In this work, we present the Self-driving Autonomous Machines for Protein Landscape Exploration (SAMPLE) platform for fully autonomous protein engineering. SAMPLE is driven by an intelligent agent that learns protein sequence-function relationships, designs new proteins, and sends designs to a fully automated robotic system that experimentally tests designed proteins and provides feedback to improve the agents understanding of the system. We deployed four SAMPLE agents with the goal of engineering glycoside hydrolase enzymes with enhanced thermal tolerance. Despite showing individual differences in their search behavior, all four agents quickly converged on thermostable enzymes that were at least 12 {degrees}C more stable than the starting sequences. Self-driving laboratories automate and accelerate the scientific discovery process and hold great potential for the fields of protein engineering and synthetic biology.

synthetic biology↗

bioRxiv2022

Inferring protein fitness landscapes from laboratory evolution experiments

Directed laboratory evolution applies iterative rounds of mutation and selection to explore the protein fitness landscape and provides rich information regarding the underlying relationships between protein sequence, structure, and function. Laboratory evolution data consist of protein sequences sampled from evolving populations over multiple generations and this data type does not fit into established supervised and unsupervised machine learning approaches. We develop a statistical learning framework that models the evolutionary process and can infer the protein fitness landscape from multiple snapshots along an evolutionary trajectory. We apply our modeling approach to dihydrofolate reductase (DHFR) laboratory evolution data and the resulting landscape parameters capture important aspects of DHFR structure and function. We use the resulting model to understand the structure of the fitness landscape and find numerous examples of epistasis but an overall global peak that is evolutionarily accessible from most starting sequences. Finally, we use the model to perform an in silico extrapolation of the DHFR laboratory evolution trajectory and computationally design proteins from future evolutionary rounds.

bioinformatics↗

bioRxiv2022

Directed evolution of angiotensin-converting enzyme 2 (ACE2) variants with enhanced peptidase activity profiles for novel therapeutic applications

Angiotensin Converting-Enzyme 2 (ACE2) is currently being investigated for its ability to beneficially modulate the Angiotensin receptor (ATR) therapeutic axis to treat multiple human diseases, but its broad substrate scope and diverse physiological roles limit its potential as a therapeutic agent. In this work we engineer ACE2 variants with enhanced hydrolytic activity and specificity toward Angiotensin-II (Ang-II) that may lead to new therapeutic candidates with improved efficacy and reduced off-target side effects. We established a yeast display-based liquid chromatography screen that enabled use of directed evolution to discover ACE2 variants with improved Ang-II activity and specificity relative to the off-target peptide substrates Apelin-13 and Angiotensin-I (Ang-I). We screened ACE2 active site libraries to reveal three substitution-tolerant positions (M360, T371 and Y510) that can be mutated to enhance ACE2s activity profile and followed up on these hits with focused double mutant libraries to further improve the enzyme. Relative to wildtype ACE2, our top variant (T371L/Y510Ile) displayed a sevenfold increase in Ang-II turnover number (kcat), a sixfold diminished catalytic efficiency (kcat/Km) on Apelin-13, and an overall decreased activity on other ACE2 substrates that were not directly assayed in directed evolution screen. At physiologically relevant substrate concentrations, T371L/Y510Ile hydrolyzes more Ang-II than wildtype ACE2 with concomitant Ang-II:Apelin-13 specificity improvements reaching 30-fold. Our efforts have delivered ATR axis-acting therapeutic candidates with relevance to both established and unexplored ACE2 therapeutic applications and provide a foundation for further ACE2 engineering efforts. Significance StatementThe Angiotensin Converting Enzyme 2 (ACE2) carboxypeptidase is being clinically trialed for treatment of acute respiratory distress syndrome (ARDS) related to traumatic injuries and viral infections. We developed an enzyme engineering platform that enabled discovery of ACE2 variants possessing enhanced peptide hydrolysis activity and specificity profiles relevant to ARDS treatment. Beyond their potential utility as ARDS therapeutics, our improved ACE2 variants could be further developed for new unexplored ACE2 therapeutic contexts such as treating Alzheimers Disease.

biochemistry↗