Engineering a Human P2X2 Receptor with Altered Ligand Selectivity in Yeast.

P2X receptors are a family of ligand gated ion channels found in a range of eukaryotic species including humans but are not naturally present in the yeast Saccharomyces cerevisiae. We demonstrate the first recombinant expression and functional gating of the P2X2 receptor in baker's yeast. We leverage the yeast host for facile genetic screens of mutant P2X2 by performing site saturation mutagenesis at residues of interest, including SNPs implicated in deafness and at residues involved in native binding. Deep mutational analysis and rounds of genetic engineering yield mutant P2X2 F303Y A304W, which has altered ligand selectivity towards the ATP analog AMP-PNP. The F303Y A304 variant shows over 100-fold increased intracellular calcium amplitudes with AMP-PNP compared to the WT receptor and has a much lower desensitization rate. Since AMP-PNP does not naturally activate P2X receptors, the F303Y A304 P2X2 may be a starting point for downstream applications in chemogenetic cellular control. Interestingly, the A304W mutation selectively destabilizes the desensitized state, which may provide a mechanistic basis for receptor opening with suboptimal agonists. The yeast system represents an inexpensive, scalable platform for ion channel characterization and engineering by circumventing the more expensive and time-consuming methodologies involving mammalian hosts.

Biosensor and machine learning-aided engineering of an amaryllidaceae enzyme.

A major challenge to achieving industry-scale biomanufacturing of therapeutic alkaloids is the slow process of biocatalyst engineering. Amaryllidaceae alkaloids, such as the Alzheimer's medication galantamine, are complex plant secondary metabolites with recognized therapeutic value. Due to their difficult synthesis they are regularly sourced by extraction and purification from the low-yielding daffodil Narcissus pseudonarcissus. Here, we propose an efficient biosensor-machine learning technology stack for biocatalyst development, which we apply to engineer an Amaryllidaceae enzyme in Escherichia coli. Directed evolution is used to develop a highly sensitive (EC50 = 20 µM) and specific biosensor for the key Amaryllidaceae alkaloid branchpoint 4'-O-methylnorbelladine. A structure-based residual neural network (MutComputeX) is subsequently developed and used to generate activity-enriched variants of a plant methyltransferase, which are rapidly screened with the biosensor. Functional enzyme variants are identified that yield a 60% improvement in product titer, 2-fold higher catalytic activity, and 3-fold lower off-product regioisomer formation. A solved crystal structure elucidates the mechanism behind key beneficial mutations.

Asymmetric Synthesis of α-Chloroamides via Photoenzymatic Hydroalkylation of Olefins.

Photoenzymatic intermolecular hydroalkylations of olefins are highly enantioselective for chiral centers formed during radical termination but poorly selective for centers set in the C-C bond-forming event. Here, we report the evolution of a flavin-dependent "ene"-reductase to catalyze the coupling of α,α-dichloroamides with alkenes to afford α-chloroamides in good yield with excellent chemo- and stereoselectivity. These products can serve as linchpins in the synthesis of pharmaceutically valuable motifs. Mechanistic studies indicate that radical formation occurs by exciting a charge-transfer complex templated by the protein. Precise control over the orientation of molecules within the charge-transfer complex potentially accounts for the observed stereoselectivity. The work expands the types of motifs that can be prepared using photoenzymatic catalysis.

Snowprint: a predictive tool for genetic biosensor discovery.

Bioengineers increasingly rely on ligand-inducible transcription regulators for chemical-responsive control of gene expression, yet the number of regulators available is limited. Novel regulators can be mined from genomes, but an inadequate understanding of their DNA specificity complicates genetic design. Here we present Snowprint, a simple yet powerful bioinformatic tool for predicting regulator:operator interactions. Benchmarking results demonstrate that Snowprint predictions are significantly similar for >45% of experimentally validated regulator:operator pairs from organisms across nine phyla and for regulators that span five distinct structural families. We then use Snowprint to design promoters for 33 previously uncharacterized regulators sourced from diverse phylogenies, of which 28 are shown to influence gene expression and 24 produce a >20-fold dynamic range. A panel of the newly repurposed regulators are then screened for response to biomanufacturing-relevant compounds, yielding new sensors for a polyketide (olivetolic acid), terpene (geraniol), steroid (ursodiol), and alkaloid (tetrahydropapaverine) with induction ratios up to 10.7-fold. Snowprint represents a unique, protein-agnostic tool that greatly facilitates the discovery of ligand-inducible transcriptional regulators for bioengineering applications. A web-accessible version of Snowprint is available at https://snowprint.groov.bio .

UTexas Aptamer Database: the collection and long-term preservation of aptamer sequence information.

A growing interest in aptamer research, as evidenced by the increase in aptamer publications over the years, has led to calls for a go-to site for aptamer information. A comprehensive, publicly available aptamer dataset, which may be a repository for aptamer data, standardize aptamer reporting, and generate opportunities to expand current research in the field, could meet such a demand. There have been several attempts to create aptamer databases; however, most have been abandoned or removed entirely from public view. Inspired by previous efforts, we have published the UTexas Aptamer Database, https://sites.utexas.edu/aptamerdatabase, which includes a publicly available aptamer dataset and a searchable database containing a subset of all aptamer data collected to date (1990-2022). The dataset contains aptamer sequences, binding and selection information. The information is regularly reviewed internally to ensure accuracy and consistency across all entries. To support the continued curation and review of aptamer sequence information, we have implemented sustaining mechanisms, including researcher training protocols, an aptamer submission form, data stored separately from the database platform, and a growing team of researchers committed to updating the database. Currently, the UTexas Aptamer Database is the largest in terms of the number of aptamer sequences with 1,443 internally reviewed aptamer records.

SARS-COV-2 Omicron variants conformationally escape a rare quaternary antibody binding mode.

The ongoing evolution of SARS-CoV-2 into more easily transmissible and infectious variants has provided unprecedented insight into mutations enabling immune escape. Understanding how these mutations affect the dynamics of antibody-antigen interactions is crucial to the development of broadly protective antibodies and vaccines. Here we report the characterization of a potent neutralizing antibody (N3-1) identified from a COVID-19 patient during the first disease wave. Cryogenic electron microscopy revealed a quaternary binding mode that enables direct interactions with all three receptor-binding domains of the spike protein trimer, resulting in extraordinary avidity and potent neutralization of all major variants of concern until the emergence of Omicron. Structure-based rational design of N3-1 mutants improved binding to all Omicron variants but only partially restored neutralization of the conformationally distinct Omicron BA.1. This study provides new insights into immune evasion through changes in spike protein dynamics and highlights considerations for future conformationally biased multivalent vaccine designs.

Using alternative SMILES representations to identify novel functional analogues in chemical similarity vector searches.

Chemical similarity searches are a widely used family of in silico methods for identifying pharmaceutical leads. These methods historically relied on structure-based comparisons to compute similarity. Here, we use a chemical language model to create a vector-based chemical search. We extend previous implementations by creating a prompt engineering strategy that utilizes two different chemical string representation algorithms: one for the query and the other for the database. We explore this method by reviewing search results from nine queries with diverse targets. We find that the method identifies molecules with similar patent-derived functionality to the query, as determined by our validated LLM-assisted patent summarization pipeline. Further, many of these functionally similar molecules have different structures and scaffolds from the query, making them unlikely to be found with traditional chemical similarity searches. This method may serve as a new tool for the discovery of novel molecular structural classes that achieve target functionality.

Parallel molecular computation on digital data stored in DNA.

DNA is an incredibly dense storage medium for digital data. However, computing on the stored information is expensive and slow, requiring rounds of sequencing, in silico computation, and DNA synthesis. Prior work on accessing and modifying data using DNA hybridization or enzymatic reactions had limited computation capabilities. Inspired by the computational power of "DNA strand displacement," we augment DNA storage with "in-memory" molecular computation using strand displacement reactions to algorithmically modify data in a parallel manner. We show programs for binary counting and Turing universal cellular automaton Rule 110, the latter of which is, in principle, capable of implementing any computer algorithm. Information is stored in the nicks of DNA, and a secondary sequence-level encoding allows high-throughput sequencing-based readout. We conducted multiple rounds of computation on 4-bit data registers, as well as random access of data (selective access and erasure). We demonstrate that large strand displacement cascades with 244 distinct strand exchanges (sequential and in parallel) can use naturally occurring DNA sequence from M13 bacteriophage without stringent sequence design, which has the potential to improve the scale of computation and decrease cost. Our work merges DNA storage and DNA computing, setting the foundation of entirely molecular algorithms for parallel manipulation of digital information preserved in DNA.

Graphene Field Effect Biosensor for Concurrent and Specific Detection of SARS-CoV-2 and Influenza.

The SARS-CoV-2 pandemic has highlighted the need for devices capable of carrying out rapid differential detection of viruses that may manifest similar physiological symptoms yet demand tailored treatment plans. Seasonal influenza may be exacerbated by COVID-19 infections, increasing the burden on healthcare systems. In this work, we demonstrate a technology based on liquid-gated graphene field-effect transistors (GFETs), for rapid and ultraprecise sensing and differentiation of influenza and SARS-CoV-2 surface protein. Most distinctively, the device consists of 4 onboard GFETs arranged in a quadruple architecture, where each quarter is functionalized individually (with either antibodies or chemically passivated control) but measured jointly. The sensor platform was tested against a range of concentrations of viral surface proteins from both viruses with the lowest tested and detected concentration at ∼50 ag/mL, or 88 zM for COVID-19 and 227 zM for Flu, which is 5-fold lower than the values reported previously on a similar platform. Unlike the classic real-time polymerase chain reaction test, which has a turnaround time of a few hours, the graphene technology presents an ultrafast response time of ∼10 s even in complex and clinically relevant media such as saliva. Thus, we have developed a multianalyte, highly sensitive, and fault-tolerant technology for rapid diagnostic of contemporary, emerging, and future pandemics.

Two sequence- and two structure-based ML models have learned different aspects of protein biochemistry.

Deep learning models are seeing increased use as methods to predict mutational effects or allowed mutations in proteins. The models commonly used for these purposes include large language models (LLMs) and 3D Convolutional Neural Networks (CNNs). These two model types have very different architectures and are commonly trained on different representations of proteins. LLMs make use of the transformer architecture and are trained purely on protein sequences whereas 3D CNNs are trained on voxelized representations of local protein structure. While comparable overall prediction accuracies have been reported for both types of models, it is not known to what extent these models make comparable specific predictions and/or generalize protein biochemistry in similar ways. Here, we perform a systematic comparison of two LLMs and two structure-based models (CNNs) and show that the different model types have distinct strengths and weaknesses. The overall prediction accuracies are largely uncorrelated between the sequence- and structure-based models. Overall, the two structure-based models are better at predicting buried aliphatic and hydrophobic residues whereas the two LLMs are better at predicting solvent-exposed polar and charged amino acids. Finally, we find that a combined model that takes the individual model predictions as input can leverage these individual model strengths and results in significantly improved overall prediction accuracy.

Analysis of macerated ticks using Boolean logic gating colorimetric isothermal nucleic acid assays for Lyme Borrelia and Ixodes scapularis ticks.

Lyme disease, one of the most common tickborne diseases, has been rapidly spreading in parallel with the expansion of the range of its tick vector. Better tick surveillance efforts are needed to accurately estimate disease risk and to guide public health and clinical management. We have developed two multiplex loop-mediated isothermal amplification (LAMP) reactions coupled with oligonucleotide strand displacement (OSD) probes to identify the tick host, Ixodes scapularis, and the Lyme disease pathogen, Borrelia burgdorferi, they carry. In each multiplex LAMP-OSD assay the co-presence of two target sequences is computed at the DNA level by linking the two corresponding amplicons and detecting the co-product on colorimetric lateral flow dipsticks. In tests with synthetic DNA, the co-presence of as few as four copies of input DNA could be detected, without producing spurious signals. Most importantly, though, the LAMP-OSD assay is amenable to being carried out directly with macerated tick samples, without any sample preparation. In such field conditions, assays performed robustly and demonstrated 97-100% sensitivity and 100% specificity with both field-collected and lab-raised artificially infected ticks. Such easy-to-use, arthropod and pathogen-specific assays would be well suited to field and near patient use without relying on complex instrumentation or infrastructure.

Peptide Macrocyclization Guided by Reversible Covalent Templating.

The creation of complementary products via templating is a hallmark feature of nucleic acid replication. Outside of nucleic acid-like molecules, the templated synthesis of a hetero-complementary copy is still rare. Herein we describe one cycle of templated synthesis that creates homomeric macrocyclic peptides guided by linear instructing strands. This strategy utilizes hydrazone formation to pre-organize peptide oligomeric monomers along the template on a solid support resin, and microwave-assisted peptide synthesis to couple monomers and cyclize the strands. With a flexible templating strand, we can alter the size of the complementary macrocycle products by increasing the length and number of the binding peptide oligomers, showing the potential to precisely tune the size of macrocyclic products. For the smaller macrocyclic peptides, the products can be released via hydrolysis and characterized by ESI-MS.

Two sequence- and two structure-based ML models have learned different aspects of protein biochemistry.

Deep learning models are seeing increased use as methods to predict mutational effects or allowed mutations in proteins. The models commonly used for these purposes include large language models (LLMs) and 3D Convolutional Neural Networks (CNNs). These two model types have very different architectures and are commonly trained on different representations of proteins. LLMs make use of the transformer architecture and are trained purely on protein sequences whereas 3D CNNs are trained on voxelized representations of local protein structure. While comparable overall prediction accuracies have been reported for both types of models, it is not known to what extent these models make comparable specific predictions and/or generalize protein biochemistry in similar ways. Here, we perform a systematic comparison of two LLMs and two structure-based models (CNNs) and show that the different model types have distinct strengths and weaknesses. The overall prediction accuracies are largely uncorrelated between the sequence- and structure-based models. Overall, the two structure-based models are better at predicting buried aliphatic and hydrophobic residues whereas the two LLMs are better at predicting solvent-exposed polar and charged amino acids. Finally, we find that a combined model that takes the individual model predictions as input can leverage these individual model strengths and results in significantly improved overall prediction accuracy.

Using machine learning to predict the effects and consequences of mutations in proteins.

Machine and deep learning approaches can leverage the increasingly available massive datasets of protein sequences, structures, and mutational effects to predict variants with improved fitness. Many different approaches are being developed, but systematic benchmarking studies indicate that even though the specifics of the machine learning algorithms matter, the more important constraint comes from the data availability and quality utilized during training. In cases where little experimental data are available, unsupervised and self-supervised pre-training with generic protein datasets can still perform well after subsequent refinement via hybrid or transfer learning approaches. Overall, recent progress in this field has been staggering, and machine learning approaches will likely play a major role in future breakthroughs in protein biochemistry and engineering.

Improved Bst DNA Polymerase Variants Derived via a Machine Learning Approach.

The DNA polymerase I from Geobacillus stearothermophilus (also known as Bst DNAP) is widely used in isothermal amplification reactions, where its strand displacement ability is prized. More robust versions of this enzyme should be enabled for diagnostic applications, especially for carrying out higher temperature reactions that might proceed more quickly. To this end, we appended a short fusion domain from the actin-binding protein villin that improved both stability and purification of the enzyme. In parallel, we have developed a machine learning algorithm that assesses the relative fit of individual amino acids to their chemical microenvironments at any position in a protein and applied this algorithm to predict sequence substitutions in Bst DNAP. The top predicted variants had greatly improved thermotolerance (heating prior to assay), and upon combination, the mutations showed additive thermostability, with denaturation temperatures up to 2.5 °C higher than the parental enzyme. The increased thermostability of the enzyme allowed faster loop-mediated isothermal amplification assays to be carried out at 73 °C, where both Bst DNAP and its improved commercial counterpart Bst 2.0 are inactivated. Overall, this is one of the first examples of the application of machine learning approaches to the thermostabilization of an enzyme.

Kinetics of elementary steps in loop-mediated isothermal amplification (LAMP) show that strand invasion during initiation is rate-limiting.

Loop-mediated isothermal amplification (LAMP) has proven to be easier to implement than PCR for point-of-care diagnostic tests. However, the underlying mechanism of LAMP is complicated and the kinetics of the major steps in LAMP have not been fully elucidated, which prevents rational improvements in assay development. Here we present our work to characterize the kinetics of the elementary steps in LAMP and show that: (i) strand invasion / initiation is the rate-limiting step in the LAMP reaction; (ii) the loop primer plays an important role in accelerating the rate of initiation and does not function solely during the exponential amplification phase and (iii) strand displacement synthesis by Bst-LF polymerase is relatively fast (125 nt/s) and processive on both linear and hairpin templates, although with some interruptions on high GC content templates. Building on these data, we were able to develop a kinetic model that relates the individual kinetic experiments to the bulk LAMP reaction. The assays developed here provide important insights into the mechanism of LAMP, and the overall model should be crucial in engineering more sensitive and faster LAMP reactions. The kinetic methods we employ should likely prove useful with other isothermal DNA amplification methods.

Mining Patents with Large Language Models Elucidates the Chemical Function Landscape.

The fundamental goal of small molecule discovery is to generate chemicals with target functionality. While this often proceeds through structure-based methods, we set out to investigate the practicality of orthogonal methods that leverage the extensive corpus of chemical literature. We hypothesize that a sufficiently large text-derived chemical function dataset would mirror the actual landscape of chemical functionality. Such a landscape would implicitly capture complex physical and biological interactions given that chemical function arises from both a molecule's structure and its interacting partners. To evaluate this hypothesis, we built a Chemical Function (CheF) dataset of patent-derived functional labels. This dataset, comprising 631K molecule-function pairs, was created using an LLM- and embedding-based method to obtain functional labels for approximately 100K molecules from their corresponding 188K unique patents. We carry out a series of analyses demonstrating that the CheF dataset contains a semantically coherent textual representation of the functional landscape congruent with chemical structural relationships, thus approximating the actual chemical function landscape. We then demonstrate that this text-based functional landscape can be leveraged to identify drugs with target functionality using a model able to predict functional profiles from structure alone. We believe that functional label-guided molecular discovery may serve as an orthogonal approach to traditional structure-based methods in the pursuit of designing novel functional molecules.

Ribosome-mediated biosynthesis of pyridazinone oligomers in vitro.

The ribosome is a macromolecular machine that catalyzes the sequence-defined polymerization of L-α-amino acids into polypeptides. The catalysis of peptide bond formation between amino acid substrates is based on entropy trapping, wherein the adjacency of transfer RNA (tRNA)-coupled acyl bonds in the P-site and the α-amino groups in the A-site aligns the substrates for coupling. The plasticity of this catalytic mechanism has been observed in both remnants of the evolution of the genetic code and modern efforts to reprogram the genetic code (e.g., ribosomal incorporation of non-canonical amino acids, ribosomal ester formation). However, the limits of ribosome-mediated polymerization are underexplored. Here, rather than peptide bonds, we demonstrate ribosome-mediated polymerization of pyridazinone bonds via a cyclocondensation reaction between activated γ-keto and α-hydrazino ester monomers. In addition, we demonstrate the ribosome-catalyzed synthesis of peptide-hybrid oligomers composed of multiple sequence-defined alternating pyridazinone linkages. Our results highlight the plasticity of the ribosome's ancient bond-formation mechanism, expand the range of non-canonical polymeric backbones that can be synthesized by the ribosome, and open the door to new applications in synthetic biology.

GroovDB: A Database of Ligand-Inducible Transcription Factors.

Genetic biosensors are integral to synthetic biology. In particular, ligand-inducible prokaryotic transcription factors are frequently used in high-throughput screening, for dynamic feedback regulation, as multilayer logic gates, and in diagnostic applications. In order to provide a curated source that users can rely on for engineering applications, we have developed GroovDB (available at https://groov.bio), a Web-accessible database of ligand-inducible transcription factors that contains all information necessary to build chemically responsive genetic circuits, including biosensor sequence, ligand, and operator data. Ligand and DNA interaction data have been verified against the literature, while an automated data curation pipeline is used to programmatically fetch metadata, structural information, and references for every entry. A custom tool to visualize the natural genetic context of biosensor entries provides potential insights into alternative ligands and systems biology.