Sequence-Based Generative AI-Guided Design of Versatile Tryptophan Synthases

Sequence-Based Generative AI-Guided Design of Versatile Tryptophan Synthases | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Sequence-Based Generative AI-Guided Design of Versatile Tryptophan Synthases View ORCID Profile Théophile Lambert , Amin Tavakoli , Gautham Dharuman , View ORCID Profile Jason Yang , Vignesh Bhethanabotla , Sukhvinder Kaur , Matthew Hill , View ORCID Profile Arvind Ramanathan , View ORCID Profile Anima Anandkumar , View ORCID Profile Frances H. Arnold doi: https://doi.org/10.1101/2025.08.30.673177 Théophile Lambert 1 Division of Chemistry and Chemical Engineering, California Institute of Technology , Pasadena, CA, USA 2 Permanent address: Université Paris-Saclay, CNRS UMR8182, Institut de Chimie Moléculaire et des Matériaux d’Orsay (ICMMO), 17 Avenue des Sciences , 91400 Orsay, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Théophile Lambert Amin Tavakoli 3 Department of Computing and Mathematical Sciences, California Institute of Technology , Pasadena, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Gautham Dharuman 4 Argonne National Laboratory , Lemont, IL, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jason Yang 1 Division of Chemistry and Chemical Engineering, California Institute of Technology , Pasadena, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jason Yang Vignesh Bhethanabotla 1 Division of Chemistry and Chemical Engineering, California Institute of Technology , Pasadena, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sukhvinder Kaur 5 Elegen Corp , 1300 Industrial Road #16, San Carlos, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Matthew Hill 5 Elegen Corp , 1300 Industrial Road #16, San Carlos, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Arvind Ramanathan 4 Argonne National Laboratory , Lemont, IL, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Arvind Ramanathan Anima Anandkumar 3 Department of Computing and Mathematical Sciences, California Institute of Technology , Pasadena, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Anima Anandkumar For correspondence: anima{at}caltech.edu frances{at}cheme.caltech.edu Frances H. Arnold 1 Division of Chemistry and Chemical Engineering, California Institute of Technology , Pasadena, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Frances H. Arnold For correspondence: anima{at}caltech.edu frances{at}cheme.caltech.edu Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract Enzymes offer unparalleled selectivity and sustainability for chemical synthesis, yet their widespread industrial application is often hindered by the slow and uncertain process of discovering and optimizing suitable biocatalysts. While directed evolution remains the gold standard for enzyme optimization, its success hinges on the availability of a starting enzyme with measurable activity, a persistent bottleneck for many desired functions. Designing libraries likely to contain such functional starting points remains a major challenge. In this work, we use the GenSLM protein language model (PLM) along with a series of filters to generate novel sequences of the β -subunit of tryptophan synthase (TrpB) that express in Escherichia coli , are stable, and are catalytically active in the absence of a TrpA partner. Many generated TrpBs also demonstrated significant substrate promiscuity, accepting non-canonical substrates typically inaccessible to natural TrpBs. Remarkably, several outperformed both natural and laboratory-optimized TrpBs on native and non-canonical substrates. Comparative analysis of the most active and promiscuous generated TrpB and its closest natural homolog confirmed that this enhanced functional versatility does not stem from the natural enzyme, highlighting the creative potential of generative models. Our results demonstrate that the model can generate enzymes which not only preserve natural structure and function but also acquire non-natural properties, establishing PLMs as powerful tools for biocatalyst discovery and engineering, with the potential in some cases to bypass further optimization. 1 Introduction Enzymes are exceptionally powerful, selective, and versatile catalysts for efficient and sustainable production of chemicals, fuels, materials, and pharmaceuticals, offering attractive alternatives to traditional chemical methods. 1 – 4 However, to meet the performance requirements of industrial applications, enzymes often require tailored optimization. In this context, directed evolution (DE) has emerged as a robust and broadly applicable strategy, using iterative rounds of mutagenesis and screening to progressively improve an enzyme toward a desired function. 5 Its success in producing industrially viable biocatalysts has been well documented, and recent advances in AI-guided design and laboratory automation can further streamline the DE workflow, significantly accelerating the pace and efficiency of enzyme engineering. 6 – 9 Despite its broad applicability, directed evolution (DE) remains fundamentally limited by the need for a starting enzyme with measurable activity toward the target function. Identifying an enzyme with such initial activity is still largely empirical, with no universal strategy currently available. One common approach is to repurpose an existing enzyme by exploiting its ability to catalyze reactions or accept substrates beyond its native biological context, a property known as catalytic or substrate promiscuity. 10 This typically begins by hypothesizing enzyme families that could accommodate the desired transformation based on mechanistic or substrate similarities. Libraries of enzymes are then constructed either by sampling natural sequence diversity or by creating mutant libraries from a small subset of representative enzymes. Both sampling natural sequences and mutating a given enzyme present challenges: natural enzymes may express poorly or have narrow specificity, while mutagenesis covers only a restricted sequence space and frequently yields a high proportion of inactive variants. The process is labor-intensive and time-consuming, with outcomes largely dictated by chance and the composition of the available enzyme libraries. Although DE has helped unlock many useful biocatalysts, 11 , 12 many promising transformations remain unexplored, and the uncertainty and duration of this process continue to limit the broader industrial adoption of enzymes as catalysts, particularly when compared to the speed and reliability of conventional synthetic chemistry. 2 , 4 To address the challenge of identifying novel enzymes with desired functions, we propose to use protein language models (PLMs) to generate libraries of enzymes that can be screened for target activities. 9 , 13 , 14 PLM-generated proteins offer significant advantages over natural ones: they can explore regions of sequence space far from known proteins, while also allowing conditioning or filtering to incorporate desirable features. 15 – 18 Recent experimental applications have validated this approach: PLMs can generate functional proteins with real-world relevance, marking a turning point for machine learning in protein engineering. 13 , 19 – 22 Building on this promise, we used the GenSLM model and developed a filtering pipeline to prioritize candidates with the highest potential for experimental validation. GenSLM was originally developed for genomescale applications, learning interactions within DNA sequences at the codon level, in contrast to most PLMs, which are trained on amino acid sequences. However, experimental validation of DNA–sequence–based models has so far been limited and remains an open area for exploration. 15 , 22 , 23 To assess the model’s generative potential, we selected a mechanistically challenging enzyme as a test case. We focused on the β -subunit of tryptophan synthase (TrpB), a subunit of the heterotetrameric tryptophan synthase complex (TrpS), a longstanding model in mechanistic enzymology. TrpS consists of two TrpA and two TrpB subunits that together catalyze a multistep biosynthetic transformation involving at least nine distinct chemical steps. TrpA produces indole, which is channeled to TrpB through a 20–25 Å substrate tunnel, where it reacts with l-serine to form l-tryptophan. This latter transformation requires the pyridoxal phosphate (PLP) cofactor and a finely tuned network of catalytic residues. Efficient catalysis depends on large-scale conformational dynamics between and within subunits, which regulate transitions between low-activity (open) and high-activity (closed) states, enabling the control of substrate binding, intermediate stabilization, and product release. The TrpB catalytic cycle is shown in Fig. 1B . Also, TrpB is an attractive industrial biocatalyst, and considerable effort has been devoted to engineering variants that function independently of TrpA, thereby simplifying their use for the synthesis of noncanonical amino acids. This endeavor began with the development of Pf TrpB-0B2 as a stand-alone catalyst and was followed by successive, labor-intensive engineering campaigns to broaden its substrate range. 24 – 26 These engineered TrpBs are now used at industrial scale: AralezBio (San Leandro, CA USA) manufactures tryptophan analogs using evolved TrpBs, and Merck & Co. (Rahway, NJ USA) employs Pf TrpB-0B2 in the synthesis of 5-fluorotryptophan, a key building block in the production of enlicitide decanoate, a phase 3 clinical candidate ( Fig. 1C ). 27 , 28 Download figure Open in new tab Figure 1. GenSLM-TrpBs: a workflow from in silico design to industrially relevant catalysts. (A) Conventional strategies for identifying starting points for directed evolution (DE) typically rely on the intrinsic promiscuity of natural enzymes and require substantial time and experimental efforts. In contrast, libraries generated by GenSLM exhibit favorable properties, including high expression, stability, and catalytic activity, as well as broadened substrate promiscuity, offering an efficient alternative for rapid biocatalyst discovery. (B) Catalytic cycle of TrpB (residue numbering based on Pf TrpB). (C) An example of industrial application of engineered TrpBs in the total synthesis of enlicitide decanoate. 27 . 2 Results 2.1 Generation of TrpBs using GenSLM The generative AI framework used here is built on the Genome-Scale Language Model (GenSLM), a large-scale transformer architecture designed to capture biological sequence patterns at the codon level. Unlike most PLMs that operate on amino acid tokens, GenSLM represents sequences as contiguous triplets of nucleotides (64 codons), directly mirroring the translation process from DNA to protein. This codon-level representation not only aligns with the central tenet of molecular biology but also enables the model to incorporate synonymous substitutions while learning protein-level effects. GenSLM was trained at multiple parameter scales—25 million (25M), 250 million (250M), 2.5 billion (2.5B), and 25 billion (25B)—using a dataset of approximately 110 million prokaryotic gene sequences from the Bacterial and Viral Bioinformatics Resource Center (BV-BRC). 29 One application of this model was to study the evolutionary dynamics of SARS-CoV-2, where fine-tuning on 1.5 million viral genomes allowed it to predict variant fitness, anticipate emerging lineages, and identify functionally relevant mutations. 23 To generate TrpB sequences, we employed the 25M parameter GenSLM and fine-tuned it on a curated dataset of trpB DNA sequences obtained from BV-BRC 29 containing 30,000 unique trpB nucleotide sequences corresponding to 22,800 unique amino acid sequences after translation. Fine-tuning followed the same procedure developed for SARS-CoV-2, using a contrastive learning objective similar to the masked language modeling strategy implemented in SpanBERT with parameters given in the Methods . 30 Given the impracticality of experimentally testing all candidates, we implemented a multi-stage in silico filtering pipeline designed to prioritize sequences that are both active and distinct from natural enzymes. Building on established sequence- and structure-based criteria, 31 the pipeline evaluates structural and sequence integrity, assesses sequence conservation with natural TrpBs, and promotes diversity across sequence space, thereby enriching the candidate pool for subsequent experimental validation. We assembled a reference database of > 57,000 natural TrpB sequences, comprising the 22,800 sequences used for fine-tuning and additional UniProt entries retrieved by querying “trpb” as the gene name and restricting the sequence length to 200–600 amino acids. This set served both as a baseline for integrity comparisons and as a diversity control in downstream filtering. Generated sequences were first filtered by length to ensure consistency with natural TrpB sequences, then modeled using ESMFold 32 ; only structures with a predicted Local Distance Difference Test (pLDDT) > 80 were retained. For each sequence, we computed its maximum sequence identity (MaxID) to the reference set and binned them as [100–90%], [90–80%], [80–70%], [70–60%], [60–50%], and [50–40%]. Sequences with MaxID > 90% were excluded to avoid selecting sequences near-identical to known TrpB enzymes and to prioritize more divergent candidates, thereby increasing the probability of identifying novel functional variants Retained sequences were assessed for sequence novelty and function, including alignment against the reference set and verifying the conservation of the catalytic lysine. Prior to final selection, we ensured that candidate sequences originated from distinct BLAST similarity clusters to avoid redundancy and to maximize coverage of diverse regions in sequence space. From this filtered set, we selected 105 representative sequences distributed as follows: 30 sequences with 80–90% MaxID, 40 with 70–80%, 20 with 60–70%, 10 with 50–60%, and 5 with 40–50%. This distribution was intentionally biased toward higher-identity sequences, which are generally associated with a greater likelihood of activity, while still preserving lower-identity sequences to explore broader sequence diversity. The final set of E. coli optimized DNA sequences, designated GenSLM-TrpBs in this article, was synthesized by Elegen Corp (Menlo Park, CA USA) and all sequences are provided in Supplementary Data 1. Embeddings from the GenSLM model were used to compare the distribution of generated sequences with that of natural TrpB sequences. The generated sequences broadly spanned the natural sequence space used for fine-tuning, as shown by the t-distributed stochastic neighbor embedding (t-SNE) plot ( Fig. 2A ). Principal component analysis (PCA) and uniform manifold approximation and projection (UMAP) analyses of the same dataset are also provided (Supplementary Fig. S2). Positional variability was highly similar between natural and generated sequences, and key conserved residues in natural TrpBs were retained in the generated sequences ( Fig. 2B-C ). Together, these results indicate that the model successfully captures the structural and evolutionary constraints that define the TrpB sequence landscape. Download figure Open in new tab Figure 2. Comparison of natural- and GenSLM-TrpBs. (A) t-SNE projection showing that GenSLM-TrpBs are well distributed across the natural TrpB sequence space used for fine-tuning. (B) Sequence variability quantified by Shannon entropy and (C) sequence logo of the most conserved residues. Both analyses indicate that GenSLM-TrpBs recapitulate natural sequence patterns. 2.2 Functional testing of GenSLM-TrpBs Our first objective was to evaluate whether the GenSLM-TrpBs exhibited activity for tryptophan synthesis. To this end, we expressed the 105 selected GenSLM-TrpBs in Escherichia coli and compared their performance to several well-characterized natural and engineered TrpB enzymes. Specifically, we selected natural TrpBs from diverse organisms available in our laboratory culture collection, which are known for their reliable heterologous expression: Escherichia coli ( Ec TrpB), Arabidopsis thaliana ( At TrpB), Pyrococcus furiosus ( Pf TrpB), Thermotoga maritima ( Tm TrpB), and Streptomyces albus ( Sa TrpB). Pf TrpB-0B2 was included as a laboratory-evolved and stand-alone TrpB. 24 Catalytic activities of the GenSLM-TrpBs and controls were evaluated at both room temperature and 75 °C. The elevated temperature condition was chosen to match the optimal temperature of Pf TrpB-0B2 and to probe the thermostability of the generated TrpBs. Residual activity from endogenous E. coli TrpS made it difficult to distinguish low-activity GenSLM-TrpBs from non-functional ones at room temperature. Despite this, 11 GenSLM-TrpBs showed activity clearly above background. Nine had 80–90% and two had 70–80% sequence identity to a natural TrpB. At 75 °C, seven GenSLM-TrpBs retained substantial activity, despite thermostability not being an explicit design criterion (five of these had 80–90% and two had 70–80% sequence identity to the closest natural TrpB). The results are shown in Fig. 3A and Supplementary Fig. S3. Download figure Open in new tab Figure 3. Tryptophan production and biophysical properties of GenSLM-TrpBs. (A) Yields of tryptophan formation catalyzed by GenSLM-TrpBs after 16 h at room temperature or 1 h at 75 °C, grouped by sequence identity to natural TrpBs and benchmarked against Pf TrpB-0B2. (B) Expression levels of the most active GenSLM-TrpBs, reported as milligrams of isolated purified protein per liter of culture. (C) Melting temperatures ( T m ) of the top GenSLM-TrpBs, measured by thermofluor assay. SeqID: sequence identity. Remarkably, several GenSLM-TrpBs exhibited levels of activity comparable to or even exceeding that of Pf TrpB-0B2, a variant of Pf TrpB specifically evolved for stand-alone function at 75 °C. 24 Given that wild-type TrpBs require activation by their TrpA subunit, the high activity of the GenSLM-TrpBs, generated without consideration of TrpA, is noteworthy. It suggests that the model can generate sequences with properties typically acquired only through extensive laboratory evolution. Among these, 230 stands out: its total activity surpasses that of Pf TrpB-0B2 at both room temperature and 75 °C. This striking result establishes that a GenSLM-designed TrpB can not only rival but even surpass a laboratory-evolved benchmark enzyme, underscoring the potential of this approach. To better characterize the most promising GenSLM-TrpBs, we purified eleven enzymes that exhibited activity at room temperature ( 1617, 2200, 2277, 2623, 3197, 3495, 3547, 3599, 3994, 230 , and 231 ), as well as one ( 1865 ) that was active at 75 °C but inactive at room temperature. Nine of these TrpBs share 80–90% sequence identity to a natural TrpB, while the remaining three share 70–80%. Expression levels were consistently high, with an average purification yield of 84 mg/L of culture, with four enzymes exceeding 100 mg/L and two approaching 200 mg/L ( Fig. 3B ). We next evaluated the thermal stability of these TrpBs. Many exhibited two melting transitions, the first between 40–50 °C, that has been associated with TrpB dimer rearrangements. 33 Melting temperatures ( T m ) were diverse: 5/12 of the tested enzymes exhibited T m above 70 °C, 6/12 were in the 50–70 °C range, and one had a T m around 40 °C ( Fig. 3C and Supplementary Fig. S5). Finally, activity assays on purified proteins conducted under varied conditions (Supplementary Fig. S4) corroborated the activity measured in 96-deep-well lysates at room temperature. However, at 75 °C, only TrpBs from thermophilic organisms ( Tm TrpB, Pf TrpB, and Pf TrpB-0B2) retained activity, whereas the GenSLM-TrpBs lost activity. This likely reflects the lower intrinsic thermostability of the GenSLM-TrpBs, which becomes apparent after purification, in contrast to the thermophilic wild-type enzymes that maintain stability under these conditions. Purification itself is known to reduce stability through factors such as buffer exposure, removal from the intracellular milieu, and a single freeze-thaw. 34 , 35 Additionally, differences in heat transfer between single-vial and plate-based assay formats could further accelerate denaturation. Interestingly, while 230 was among the most active in lysate, its performance decreased following purification. In contrast, 3599, 3994 , and 3547 exhibited superior activity in the purified format, achieving near-quantitative conversions under shorter reaction times and reduced catalyst loadings. 2.3 Substrate promiscuity of GenSLM-TrpBs We hypothesized that TrpB generated via a PLM might exhibit broader substrate promiscuity compared with their counterparts that evolved naturally. This hypothesis stems from the observation that natural enzymes like TrpB are typically highly specific, having undergone millions of years of evolutionary pressure to suppress promiscuous activities that could disrupt cellular homeostasis. 10 , 36 PLM-generated (or other computationally designed) enzymes may be less optimized and inherently more permissive, potentially offering greater substrate flexibility. Enzymes reconstructed through Ancestral Sequence Reconstruction (ASR), for example, have been shown to display a broader substrate range. 37 While this has often been attributed to the re-creation of a more permissive ancestral state, it remains unclear whether the observed promiscuity arises from genuine ancestral features or from biases introduced by the reconstruction method itself. To investigate this, all 105 GenSLM-TrpBs were screened for activity on a panel of non-cognate substrates using lysates in a 96-deep-well format. While natural tryptophan synthase (TrpS) enzymes can accept substituted indoles, the activity is typically limited, and their substrate range is narrow. Reported indole substitutions accepted by the native enzymes include halogens and electron-donating substituents such as methyl, amino, methoxy, or hydroxy groups. 38 – 41 To challenge the generated TrpBs, we selected seven substrates that are poorly reactive with wild-type TrpBs and therefore have been targeted in prior directed evolution studies. These included 4-NO 2 -, 5-NO 2 -, 6-CN- and 7-CN-indole. Naphthol is a non-indole compound that no natural TrpS has been reported to process. 42 l-threonine was also tested as an alternative electrophile, as natural TrpS enzymes display strict specificity for l-serine. 43 Finally, 5-fluoroindole was included due to its relevance in the industrial synthesis of enlicitide decanoate ( Fig. 1C ), although it is known to be accepted by natural TrpBs. 27 , 40 . We compared the generated TrpBs against enzymes that were specifically evolved in the laboratory for different activities: Pf TrpB-0B2 24 ( Pf TrpB stand-alone), Tm Triple 44 ( Tm TrpB stand-alone), Tm 9D8* 45 (activity on 4-CN-indole at lower temperatures), Pf 0A9 and Pf 2A6 25 (activity on 4-NO 2 -indole and indole derivatives), Pf 2B9 43 (activity with l-threonine), and Tm TyrS6/ Tm 9D8* E105G 42 (activity for phenol/naphthol to make tyrosine derivatives). Wild-type TrpBs from five species ( Ec TrpB, At TrpB, Pf TrpB, Tm TrpB, and Sa TrpB) were included for comparison. Product identity was confirmed by comparison to authentic standards, and reaction yields were estimated at the isosbestic point (277 nm). 46 The results are shown in Fig. 4B , with the complete data available in the Supplementary Fig. S6. Download figure Open in new tab Figure 4. Reaction yields of GenSLM-TrpBs with non-natural substrates. (A) Structures of non-canonical substrates tested. (B) Product yields estimated by UV absorbance at the isosbestic point (277 nm). Yellow dashed boxes denote reactions where product formation was confirmed by mass spectrometry above background levels, but remained below the UV detection threshold. Yields are displayed using a power-law normalization with γ = 0.15. (C) Number of substrates accepted per TrpB, comparing GenSLM-TrpBs (clustered by sequence identity) with evolved and natural TrpBs. (D) Comparative yields of the two most promiscuous GenSLM-TrpBs compared to the most promiscuous natural enzyme ( At TrpB), the most promiscuous evolved TrpB ( Pf 2B9), and the industrially relevant Pf TrpB-0B2. SeqID: sequence identity. Strikingly, for every substrate tested, at least one GenSLM-TrpB exhibited measurable activity. In particular, GenSLM-TrpBs with higher sequence identity to natural TrpBs (70–80% and 80–90%) contained enzymes that showed greater promiscuity compared to the natural TrpBs, while the laboratory-evolved TrpBs consistently displayed high promiscuity ( Fig. 4C ). A few GenSLM-TrpBs were active on the most challenging substrates — 4-NO 2 -indole, naphthol, and threonine — where only 230 consistently produced UV-detectable product. Nevertheless, some additional GenSLM-TrpBs showed detectable activity by mass spectrometry, providing viable starting points for directed evolution. In contrast, a broader range of active enzymes was identified for 5-NO 2 -indole, 6-CN-indole, and 7-CN-indole, with many outperforming the natural TrpBs. For 5-fluoroindole, a high background activity from endogenous E. coli TrpS was observed. Yet several GenSLM-designed enzymes such as 230, 1617 , and 3599 achieved impressive yields (99%, 97%, and 60% respectively), significantly exceeding the performance of natural TrpBs and almost reaching the quantitative yield of Pf TrpB-0B2, which is used industrially for this exact reaction. Among these, 230 was particularly notable; it exhibits measurable activity across all tested substrates with yields ranging from 5% to 99%. This degree of substrate promiscuity is unprecedented among natural TrpBs. At 37 °C, 230 matched or exceeded the performance of the laboratory-evolved enzymes in reactions with 4-NO 2 -indole, 5-NO 2 -indole, l-threonine, and 7-CN-indole. 3599 exhibited detectable activity across the entire substrate panel, though at lower levels than 230. 1617 was active on six of the seven tested substrates. Figure 4D compares the product yields of these two most promiscuous GenSLM-TrpBs with those of the most promiscuous natural enzyme ( At TrpB), the evolved Pf 2B9, and the industrially relevant Pf TrpB-0B2. 2.4 GenSLM introduces functional improvements beyond natural sequence diversity To better understand the origin of the properties observed in the GenSLM-TrpBs, we asked a central question: do these features merely reflect the natural sequence distribution captured by the model, or does the model introduce novel attributes beyond what is found in nature? We previously demonstrated that GenSLM-generated sequences broadly span the natural TrpB sequence space while preserving key structural and evolutionary constraints. Notably, catalytically active GenSLM-TrpBs are evenly distributed across this space, indicating that the model sampled multiple distinct functional solutions rather than converging on a single sequence cluster (Supplementary Fig. S7). Despite this close identity with natural sequences, GenSLM-TrpBs exhibited enhanced substrate promiscuity relative to all tested natural homologs. As testing the promiscuity of all the 57,000 known natural TrpBs is impractical, we instead focused on a single representative GenSLM-TrpB, 230 , which exhibited both high catalytic activity and broad substrate range. To assess the promiscuity of 230 compared to natural TrpBs, we looked at its closest natural homolog, TrpB from Neobacillus drentensis ( Nd TrpB, NCBI ID: WP 335697934.1), which shares 80.5% sequence identity (322 of 400 residues) and had not been previously characterized. Structural modeling with ESMFold 32 predicted nearly identical folds, with a backbone RMSD of 0.87 Å( Fig. 5C ). The gene encoding Nd TrpB was synthesized, expressed in E. coli , and the enzyme was purified alongside 230 using the same methods. Nd TrpB was expressed at a higher level (75 mg/L of culture) than 230 (5 mg/L), but they exhibited comparable thermal stability ( T m = 76.5 °C for Nd TrpB and T m = 77.5 °C for 230 ). Catalytically, however, they differed substantially. Using E. coli lysate at room temperature, both enzymes showed high product yields (94% for Nd TrpB and 92% for 230 ). However, at 75 °C, Nd TrpB activity dropped sharply to 19%, whereas 230 retained nearly full activity (94%). More strikingly, 230 exhibited significantly broader reactivity across a panel of non-canonical substrates, outperforming Nd TrpB in every case. Nd TrpB showed detectable activity only with 7-cyanoindole and 5-fluoroindole, substrates previously known to be accepted by natural TrpBs 26 , and failed to produce any product with more challenging substrates ( Fig. 5A ). Download figure Open in new tab Figure 5. Comparison of GenSLM-TrpB 230 and its closest natural homolog from Neobacillus drentensis ( Nd TrpB). (A) Product yields with the native substrate at room temperature and 75 °C, as well as with various non-canonical substrates. (B) Comparison of melting temperatures and expression levels (mg/L). (C) Structural alignment predicted by ESMFold (backbone RMSD = 0.87 Å); 230 is shown in orange, Nd TrpB in green, and non-conserved residues highlighted in purple. 3 Discussion Protein engineering is undergoing a major transformation driven by advances in artificial intelligence (AI), which is reshaping how enzymes are designed and optimized. AI models trained on large biological datasets can generate enzymes with remarkable diversity and functional performance. 15 , 47 , 48 While de novo protein design enables the creation of proteins able to catalyze non-natural reactions, 49 , 50 this approach remains limited to relatively simple reactions and is currently not applicable to highly complex systems like TrpB, whose mechanism involves two substrates, a cofactor, dynamic conformational changes, and a multistep catalytic cycle. Nevertheless, our results show that GenSLM-TrpBs were expressible, catalytically competent, stable, and broadly promiscuous. Remarkably, several GenSLM-TrpBs outperformed both natural and laboratory-evolved TrpBs on the native indole substrate, as well as on non-canonical substrates. Among the designs tested, 230 emerged as an extraordinary example. This enzyme outperforms the extensively engineered Pf TrpB-0B2 in tryptophan formation at both room temperature and 75 °C, while achieving comparable yield for the synthesis of 5-fluorotryptophan. Given Pf TrpB-0B2’s industrial relevance and its long history of optimization for stand-alone activity, the discovery of a superior, PLM-generated enzyme was both surprising and exciting. Beyond its exceptional catalytic activity, 230 also displays unprecedented substrate promiscuity, catalyzing all tested non-native reactions, a property not observed in the natural TrpBs tested. Direct comparison between 230 and its closest natural homolog ( Nd TrpB) showed that, although both enzymes display similar activity at room temperature, the natural homolog lacks high-temperature stability and broad substrate scope. This confirms that the versatility of 230 cannot be explained as simply reproducing a natural enzyme with similar properties. The enhanced activity and substrate promiscuity of GenSLM-TrpBs cannot be attributed simply to their divergence from natural sequence space. Instead, they fall within the distribution of natural TrpBs, retaining catalytic activity while exhibiting properties that differ from their natural homologs. This mirrors the concept of neutral drift in protein evolution, whereby proteins accumulate mutations while maintaining their original structure and function. As a result, GenSLM-TrpBs resemble laboratory-drifted variants and, like neutrally drifted proteins, can acquire new properties relative to their natural counterparts including enhanced substrate promiscuity, a phenomenon well documented in previous studies. 46 , 51 – 53 Thus, GenSLM-generated enzymes may represent a practical route to exploit this evolutionary mechanism, yielding drifted scaffolds that preserve natural function while exhibiting expanded promiscuity. Previous studies have shown that the internal representations learned by PLMs capture meaningful aspects of biological structure and function. 21 , 22 , 54 , 55 While much effort is focused on refining these models to closely replicate natural enzymes and their functions, we argue that their inability to fully do so is not a limitation, but rather a strength. This divergence from nature provides a unique opportunity to explore sequence space beyond evolutionary constraints and uncover novel biocatalysts with new or enhanced properties. In essence, we believe that by freeing itself from biological constraints, generative AI offers the perfect foundations for biocatalyst discovery . 4 Conclusion These results underscore the potential of AI-driven design for making new and useful enzymes. In an optimized pipeline, the entire cycle of design, gene synthesis, and experimental testing can be completed in as little as one month, providing an exceptionally rapid route to functional enzymes. Our work shows that the GenSLM-generated enzyme library combines high activity with broad substrate scope making it an an ideal starting point for the exploration of new substrates and the evolution of new enzymatic functions while reducing experimental burden. Moreover, the fact that some generated enzymes already rival or even surpass laboratory-evolved enzymes suggests that, in some cases, it could eliminate the need for directed evolution altogether, offering a formidable acceleration for biocatalyst design, a major bottleneck limiting biocatalyst use at an industrial scale. Author contributions T.L. conceived the experimental study, designed its methodology, performed the wet-lab experiments, wrote the manuscript, and combined the editing. A.T. was in charge of the machine learning, including model fine-tuning, data collection and filtering pipeline, and helped with editing. G.D. contributed to model fine-tuning and data collection. J.Y. provided general advice, performed quality control and helped with editing. V.B. helped in the design of the filtering pipeline. S.K. and M.H. coordinated DNA synthesis. A.R. provided resources, supervision, and funding. A.A. contributed resources, manuscript editing, supervision, and funding. F.H.A. oversaw the project, provided resources and funding, and contributed to manuscript editing and supervision. Code Availability All data and code used will be made publicly available. This repository will include the reference TrpB sequences, the fine-tuning dataset, the final selected candidate sequences, the pretrained GenSLM model, the fine-tuned GenSLM checkpoints, and the code implementing the computational filtering pipeline. Competing Interest The authors declare no competing interests. Methods General experimental methods All chemicals were obtained from commercial suppliers and used without further purification. Analytical liquid chromatography–mass spectrometry (LC–MS) was performed using an Agilent 1260 Infinity II LC/MSD-iQ system. PCR reactions were carried out on an Eppendorf Mastercycler X50s. The 96-well deep-well plates were shaken unsing an INFORS HT Multitron Shaker at 220 rpm, 80% humidity at the given temperature. Isolation of plasmids were realized using the Monarch Miniprep Kit (NEB, Ipswich, MA) according to the manufacturer’s protocol. All enzymes in this study contain a C-terminal 6×His tag to enable affinity purification. Cloning and transformation The pET22b(+) vector was linearized by inverse PCR using Phusion polymerase (NEB, Ipswich, MA) and primers 007 and 008 (008 Forward: CTCGAGCACCACCACCACCACCACTGAGATCCGGC; 007 Reverse: CATATGTATATCTCCTTCTTAAAGTTAAACAAAATTATTTC) digested with DpnI (NEB, Ipswich, MA), gel-purified, and validated to ensure minimal background transformation. A total of 95 synthetic genes were synthesized by Elegen Corp. (San Carlos, CA). DNA fragments containing flanking regions (upstream: GTT-TAACTTTAAGAAGGAGATATACAT; downstream: CTCGAGCACCACCATCACCACCACTGA) for Gibson assembly were received as dry residues in 96-well plates at approximately 2 μ g per well. These DNA samples were dissolved in PCR-grade water to a concentration of approximately 50 ng/ μ L. Following NEB recommendations, 1 μ L of DNA fragment was mixed with 0.5 μ L of linearized pET22b(+) (200 ng/ μ L) and 5 μ L of Gibson assembly mix in a 96-well PCR plate. The plate was sealed and incubated at 50 °C for 60 min, then placed on ice. Subsequently, 5 μ L of chemically competent E. coli T7 Express cells (NEB, Ipswich, MA) were added to each well, followed by a 20 min incubation on ice and a 10 s heat shock in a 42 °C water bath. After transformation, 100 μ L of Luria–Bertani (LB) medium were added to each well, and 10 μ L of the mixture were used to inoculate 500 μ L of LB containing 100 μ g/mL ampicillin (LB amp ) in a 96-deep-well plate and incubated at 37 °C for 16–18 h. Cultures were passaged once by reinoculating 10 μ L into fresh 500 μ L LB amp . Successful transformants were verified via in-house sequencing (LevSeq). 56 Additional genes (10 GenSLM-TrpBs, Nd TrpB) were synthesized by Twist Bioscience (South San Francisco, CA) and processed following a similar protocol, with the exception that transformed cells were plated on LB amp agar instead of being inoculated into liquid media. Plasmids were isolated from selected colonies and sent for sequencing (Transnetyx, Inc., Cordova, TN). Validated plasmids were transformed into E. coli T7 Express, and glycerol stocks were prepared for long-term storage. Plasmids encoding control and previously characterized variants were retrieved from our in-house collection and transformed using the same procedure. For screening native activity, two 96-well plates were assembled containing all GenSLM-TrpBs along with control constructs. These included wild-type TrpBs ( At TrpB, Pf TrpB, Tm TrpB, and Sa TrpB), the evolved variant Pf TrpB-0B2, the empty vector pUC19 (negative control), and a sterile well. For promiscuity screening, two additional plates were assembled by supplementing the initial set with previously engineered TrpB variants: Tm Triple 44 , Tm 9D8* 45 , Pf 0A9 25 , Pf 2A6 25 , Pf 2B9 43 , Tm TyrS6 42 , and Tm 9D8* E105G 42 . Control variants were included on each plate to ensure consistent comparisons across the different plates. Glycerol stocks of all transformants were prepared and archived for long-term storage and for future inoculation. Analytical scale analysis in 96-well plate Glycerol stocks were used to inoculate 300 μ L LB amp in 96-well plates, covered with a sterile, breathable film, and grown at 37 °C overnight. From the stationary-phase cultures, 50 μ L were transferred into 900 μ L TB containing 100 μ g/mL ampicillin (TB amp ) and incubated for 2 h at 37 °C prior to induction with with 50 μ L of IPTG in TB amp (0.5 mM final). Induced cultures were incubated for 22 h at 22 °C to allow protein expression. Cells were then harvested by centrifugation (4,000× g , 5 min), and the resulting pellets were either processed immediately or stored at −20 °C for later use. All reactions were run using E. coli lysate. For lysate preparation in 96-well plates, cell pellets were resuspended in 500 μ L of lysis buffer (100 mM KPi, pH 8.0, supplemented with 100 μ M PLP, 1 mg/mL lysozyme, 2 mM Mg 2+ , and DNase I) and incubated at 37 °C for 1 h with shaking at 200 rpm. Lysis was completed via three freeze–thaw cycles (≥5 min in an ethanol/dry ice bath, > 30 min thaw at room temperature, followed by > 5 min in a 37 °C water bath). Cell debris was removed by centrifugation (6,000× g , 15 min), and 300 μ L of clarified lysate were transferred to fresh 96-deep-well plates. For the canonical l-tryptophan synthesis assay, 10 mM l-serine were added and the volume was adjusted to reach 390 μ L using 100 mM KPi buffer. Reactions were initiated by adding 10 μ L of indole (400 mM in ethanol; final cosolvent concentration 2.5%). All assays were performed in biological duplicate and incubated either at room temperature for 16 h or at 75 °C for 1 h in a pre-warmed water bath. Reactions were quenched by addition of 400 μ L acetonitrile (MeCN), vortexed, and centrifuged at 6,000× g for 10 min. From the supernatant, 300 μ L were mixed with 600 μ L MeCN, clarified again by centrifugation, and 200 μ L were transferred to 96-well Agilent plates for LC-MS analysis. The final 6-fold dilution in MeCN ensured the removal of salts, proteins, and particulate matter prior to injection. For promiscuity screening, 10 mM l-serine or l-threonine were added and the volume was adjusted to reach 340 μ L using 100 mM KPi buffer. Reactions were initiated by the addition of 10 μ L of indole derivatives (350 mM in EtOH or DMSO; final cosolvent concentration 2.9%). Plates were incubated for 20 h at 37 °C in a Kuhner Shaking incubator at 160 rpm. Reactions were quenched with 700 μ L of a 3:1 mixture of MeCN and 1 M HCl, vortexed, and centrifuged (6,000× g , 10 min). A 300 μ L aliquot of the supernatant was diluted with 600 μ L MeCN, enabling a final 6-fold dilution in MeCN, clarified by a second centrifugation, and 200 μ L were transferred to 96-well Agilent plates for LC-MS analysis. LC-MS screening The worked-up samples were transferred into assay plates, sealed, and analyzed by LC-MS using a reversed-phase Poroshell 120 EC-C18 column (4.6 × 50 mm, 2.7 μ m) equipped with a C18 guard column. The chromatographic method employed a solvent system of H 2 O/MeCN with 0.1% acetic acid at a flow rate of 1 mL/min. The gradient started at 5% MeCN for 0.5 min, increased linearly to 95% MeCN over 1.5 min, held at 95% for 0.7 min, then decreased back to 5% MeCN in 0.3 min, followed by a 1-min post-run equilibration. Yields for canonical reactions were quantified by UV absorbance at 277 nm, corresponding to the isosbestic point between indole and tryptophan. For promiscuity screening, standard products were prepared from evolved variants of previous campaigns, and yields were similarly estimated at 277 nm, as described previously 46 . When product levels were below the UV detection limit, products were considered detected only if the MS signal exceeded three times the standard deviation of the negative control (pUC19), ensuring rigorous exclusion of background noise. Protein purification For large-scale expression, single colonies were inoculated into LB amp and grown overnight. The following day, TB amp media were inoculated at a 1:100 dilution from the overnight culture and incubated at 37 °C until the culture reached an OD 600 of 0.6–0.8. Protein expression was induced with 0.5 mM IPTG and allowed to proceed at 22 °C for 20–22 hours. Cells were harvested by centrifugation (5,000× g , 10 min) and stored at −20 °C until further use. Lysates were clarified by centrifugation ( > 15,000× g , 30 min) and loaded onto 1-mL HisTrap columns using an AKTA Xpress system preequilibrated with buffer A (50 mM KPi, 200 mM NaCl, 20 mM imidazole, pH 8.0). Columns were washed with 10 column volumes (CV) of buffer A and proteins were eluted using a gradient to buffer B (50 mM KPi, 200 mM NaCl, 400 mM imidazole, pH 8.0). Eluted proteins were buffer-exchanged by dialysis into 100 mM KPi pH 8.0, aliquoted, flash-frozen in liquid nitrogen, and stored at −80 °C. Each aliquot was thawed only once and used on the same day. Protein concentrations were determined by absorbance at 280 nm using predicted extinction coefficients and molecular weights. Protein purity was evaluated by SDS–PAGE (Mini-PROTEAN TGX, 4–20%) using the Precision Plus Protein™ Kaleidoscope™ ladder, following the manufacturer’s instructions. Analytical-scale analysis with pure protein Proteins were thawed on ice and normalized to the specified concentration. Pyridoxal phosphate (PLP) was added at a fivefold molar excess relative to protein. Reactions were prepared in 200 μ L volumes in triplicate in Eppendorf tubes, containing 10 mM indole, 20 mM serine, and 5% ethanol as co-solvent. For reactions at 75 °C, tubes were incubated in a pre-warmed water bath, while those at 25 °C were incubated in a Kuhner shaker incubator at 160 rpm. Upon completion, 1 mL of MeCN was added to quench the reactions, followed by vortexing and centrifugation at maximum speed for 10 minutes. The clarified supernatant was then used for LC-MS analysis. Analytical-scale analysis with lysate To compare the activity of 230 and Nd TrpB, fresh transformants were inoculated into 5 mL LB amp and grown overnight. The following day, 50 mL TB amp cultures were inoculated at a 1:100 dilution from the overnight culture and grown at 37 °C until reaching an OD of 0.6–0.8. Protein expression was induced with 0.5 mM IPTG and continued at 22 °C for 20–22 hours. Cells were harvested by centrifugation (5,000× g , 10 min) and stored at −20 °C. Cell pellets were resuspended in lysis buffer at an OD of 30 (100 mM KPi, pH 8.0, 100 μ M PLP, DNase I, 1 mg/mL lysozyme, 2 mM Mg 2+ ), incubated at 37 °C for 1 hour, and lysed by sonication (2 min, 1 s on/1 s off, 35% amplitude). For each reaction, 300 μ L of lysate were mixed with the electrophile (l-serine or l-threonine) and the corresponding indole derivative dissolved in DMSO or ethanol to a final cosolvent concentration of 2.5%. Reactions were performed in triplicate and processed as described in the Analytical scale analysis in 96-well plate section. Melting temperature determination Melting temperatures ( T m ) were determined using a thermofluor assay with SYPRO Orange on a Bio-Rad CFX96 Touch Real-Time PCR system. Purified enzymes were normalized to 5 μ M and supplemented with 10 μ M PLP. A total volume of 150 μ L of the protein–PLP mixture was combined with 10 μ L SYPRO Orange (200X stock). Three 50 μ L aliquots were transferred into a qPCR 96-well plate. The temperature was increased from 25 to 99 °C in 0.5 °C increments, holding each step for 30 seconds with fluorescence measurements taken at the end of each interval. Comparison between generated and natural TrpBs GenSLM embeddings of both natural and generated TrpB sequences were used to create dimensionality reduction plots. Principal Component Analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE) were performed using Scikit-Learn. Uniform Manifold Approximation and Projection (UMAP) visualizations were generated with the python umap package, employing parameters n neighbors=15 and min dist=0.5, while all other settings were left at their defaults. To assess positional variability, natural and generated sequences were aligned using MAFFT. High-gap regions were trimmed to produce a final alignment length of 408 amino acids, corresponding to the average length of natural TrpBs. Positional variability was quantified using Shannon entropy, and sequence logos were generated from this alignment using Logomaker (Bioconda). Fine-tuning The TrpB sequences were retrieved from the Bacterial and Viral Bioinformatics Resource Center (BV-BRC) database using EC classification number 4.2.1.20. Low quality sequences were filtered out by keeping only sequences coming from complete genomes with less than 5% contamination and that contain less than 50 contigs. Since the EC class matches both the alpha and beta subunits, only features annotated as the beta subunit were included in the final dataset. Fine-tuning was performed for 10 epochs with full parameter updates, using the standard autoregressive cross-entropy objective at the codon level. No diffusion-based hierarchical modeling or reward-guided generation was used in this stage. Following fine-tuning, the model was used to generate a batch of 10,000 trpB gene sequences, each prompted with the methionine start codon (ATG). Sequence generation employed nucleus sampling (top p = 0.9) in combination with top-k filtering (top k = 50), with temperature set to 1.0. Generation was bounded to a maximum of 512 codon tokens, with early termination if a designated stop token was encountered. Sequences were generated in batches of 512. Filtering pipeline A schematic depiction of the filters is shown in SI (Fig. S1). It is important to note that the order of filters is determined to maximize the novelty and functionality of the filtered sequences. The fine-tuned GenSLM generates trpB codon sequences in an autoregressive manner, conditioned on an initial token provided as a prompt. To streamline the downstream filtering process, all generated nucleotide sequences were translated into their corresponding amino acid sequences. Start Codon Filter We first filtered sequences to retain only those that begin with the canonical start token as the ATG codon. This step ensures translational initiation compatibility and reflects biological realism. All trpB -prompted sequences passed this filter. Length Filter Next, we enforced a length constraint based on the natural trpB sequence distribution (references). Sequences were accepted if their length L satisfied: where ℒ = 363.55 and σ = 57.91 – statistics obtained from the references, result in a valid range of 306 to 421 amino acids. This filter ensures the resulting sequences remain within a functional length range typical of the protein family. Among trpB -prompted sequences, 89% passed the length filter. Structural Integrity Filter We then assessed structural integrity using ESMFold. 32 Only sequences with predicted Local Distance Difference Test (pLDDT) scores ≥80% were retained, indicating high folding confidence of the protein backbone. This filter removed sequences unlikely to adopt stable structures. Of the remaining length-filtered sequences, 93% passed the structural confidence threshold. The autoregressive nature of GenSLM makes the initial token critical, as it influences downstream sequence generation. Additionally, maintaining sequence lengths similar to natural variants is essential, as protein function and folding are often length-dependent within a given family. Applying these filters to a batch of 5,000 generated TrpBs resulted in pass rates of 95% for length and 90% for structural confidence (Fig. S1B). Optional Filter 1: Stability Filter We also evaluated Rosetta energy scores 57 as a proxy for thermodynamic stability. Although this metric can be used to remove unstable sequences, we did not apply a hard threshold in this round of selection. First, folding confidence (pLDDT) and Rosetta energy are correlated, so unstable sequences are already partly filtered. Second, strictly filtering by stability may reduce the novelty of the generated proteins and bias selection toward variants that function only under specific conditions, thus limiting potential discovery. Sequence Novelty via Max Identity To assess novelty, we computed the maximum sequence identity (MaxID) of each filtered sequence relative to a reference set of natural TrpBs, following REF 19 . Figure S1C shows the distribution of MaxID values, where it peaks near 100% identity, reflecting the high fidelity of the generative model and the stringency of the initial filters. Partitioning by Sequence Identity To systematically explore sequence novelty, we partitioned the filtered sequences into six identity bins based on max ID: (40–50%), (50–60%), (60–70%), (70–80%), (80– 90%), and (90–100%). This partitioning enables balanced sampling of sequences across different similarity levels while retaining plausible function. Figure S1C shows the population size within each bin. Ranking by Alignment Score Within each bin, we prioritized sequences using a custom Alignment Score , defined as a weighted sum of global and local pairwise alignment scores to all reference sequences. Grid search identified optimal weights: 3 for global and 8 for local alignment. Scoring used +1 for residue matches, −1 for mismatches, −0.5 for gap openings, and −0.1 for gap extensions. Sequences were ranked accordingly, and the top candidates from each bin were selected. Optional Filter 2: Active Site Conservation To preserve catalytic functionality, we aligned all candidates to Pf TrpB (PDB: 5dw0) using multiple sequence alignment and retained only sequences with a conserved lysine at position 82, the known catalytic residue. Optional Filter 3: Sequence Diversity To ensure diversity among final candidates, we clustered sequences using BLAST at a 70% sequence similarity threshold and selected only one representative per cluster. This step promotes coverage of distinct sequence families and reduces redundancy. Final Selection All sequences with > 90% identity to any natural variant were excluded. From the remaining identity bins, we selected 30 sequences with 80–90% MaxID, 40 with 70–80%, 20 with 60–70%, 10 with 50–60%, and 5 with 40–50%, totaling 105 sequences. All sequences were codon-optimized for E. coli expression. The ordered DNA sequences are provided in Supplementary Data 1. Acknowledgments This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award Number DE-SC0022218. This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. T.L. gratefully acknowledges financial support for this research by the Fulbright Program, which is sponsored by the U.S. Department of State and the Franco-American Commission – Fulbright France. Its contents are solely the responsibility of the author and do not necessarily represent the official views of the Fulbright Program, the Government of the United States, or the Franco-American Commission. A.A. is supported by the Bren endowed chair and the Schmidt AI2050 senior fellowship. The authors thank Sabine Brinkmann-Chen for critical reading of the manuscript. The authors also thank Kyle Hippe for helping with extracting and filtering the TrpB sequences and Ariane Mora for checking the integrity of the sequences. Funder Information Declared U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences , DE-SC0022218 Fulbright France Schmidt AI2050 References 1. ↵ Winkler , C. K. , Schrittwieser , J. H. & Kroutil , W. Power of biocatalysis for organic synthesis . ACS Central Science 7 , 55 – 71 ( 2021 ). OpenUrl PubMed 2. ↵ Truppo , M. D. Biocatalysis in the pharmaceutical industry: the need for speed . ACS Medicinal Chemistry Letters 8 , 476 – 480 ( 2017 ). OpenUrl PubMed 3. Huisman , G. in Comprehensive Organic Synthesis (Second Edition) 421 – 437 ( Elsevier , 2014 ). isbn: 978-0-08-097743-0 . 4. ↵ Adams , J. P. , Brown , M. J. , Diaz-Rodriguez , A. , Lloyd , R. C. & Roiban , G.-D. Biocatalysis: A pharma perspective . Advanced Synthesis & Catalysis 361 , 2421 – 2432 ( 2019 ). OpenUrl 5. ↵ Wu , S. , Snajdrova , R. , Moore , J. C. , Baldenius , K. & Bornscheuer , U. T. Biocatalysis: enzymatic synthesis for industrial applications . Angewandte Chemie International Edition 60 , 88 – 119 ( 2021 ). OpenUrl PubMed 6. ↵ Zhong , Z. et al. Automated continuous evolution of proteins in vivo . ACS Synthetic Biology 9 , 1270 – 1276 ( 2020 ). OpenUrl PubMed 7. Yu , T. , Boob , A. G. , Singh , N. , Su , Y. & Zhao , H. In vitro continuous protein evolution empowered by machine learning and automation . Cell Systems 14 , 633 – 644 ( 2023 ). OpenUrl PubMed 8. Yang , J. et al. Active learning-assisted directed evolution . Nature Communications 16 , 714 ( 2025 ). OpenUrl PubMed 9. ↵ Yang , J. Li , F.-Z. & Arnold , F. H. Opportunities and challenges for machine learning-assisted enzyme engineering . ACS Central Science 10 , 226 – 241 ( 2024 ). OpenUrl PubMed 10. ↵ Tawfik , O. K. & S, D. Enzyme promiscuity: a mechanistic and evolutionary perspective . Annual Review of Biochemistry 79 , 471 – 505 ( 2010 ). OpenUrl CrossRef PubMed Web of Science 11. ↵ Reisenbauer , J. C. , Sicinski , K. M. & Arnold , F. H. Catalyzing the future: recent advances in chemical synthesis using enzymes . Current Opinion in Chemical Biology 83 , 102536 ( 2024 ). OpenUrl CrossRef PubMed 12. ↵ Chen , K. & Arnold , F. H. Engineering new catalytic activities in enzymes . Nature Catalysis 3 , 203 – 213 ( 2020 ). OpenUrl 13. ↵ Ruffolo , J. A. & Madani , A. Designing proteins with language models . Nature Biotechnology 42 , 200 – 202 ( 2024 ). OpenUrl CrossRef PubMed 14. ↵ Ferruz , N. & Höcker , B. Controllable protein design with language models . Nature Machine Intelligence 4 , 521 – 532 ( 2022 ). OpenUrl 15. ↵ Albanese , K. I. , Barbe , S. , Tagami , S. , Woolfson , D. N. & Schiex , T. Computational protein design . Nature Reviews Methods Primers 5 , 13 ( 2025 ). OpenUrl 16. Nijkamp , E. , Ruffolo , J. A. , Weinstein , E. N. , Naik , N. & Madani , A. Progen2: exploring the boundaries of protein language models . Cell Systems 14 , 968 – 978 ( 2023 ). OpenUrl CrossRef PubMed 17. Ferruz , N. , Schmidt , S. & Höcker , B. ProtGPT2 is a deep unsupervised language model for protein design . Nature Communications 13 , 4348 ( 2022 ). OpenUrl PubMed 18. ↵ Yang , J. , Bhatnagar , A. , Ruffolo , J. A. & Madani , A. Function-Guided Conditional Generation Using Protein Language Models with Adapters . Preprint at https://arxiv.org/abs/2410.03634 ( 2025 ). 19. ↵ Madani , A. et al. Large language models generate functional protein sequences across diverse families . Nature Biotechnology 41 , 1099 – 1106 ( 2023 ). OpenUrl CrossRef PubMed 20. Munsamy , G. et al. Conditional language models enable the efficient design of proficient enzymes . Preprint at https://www.biorxiv.org/content/early/2024/05/05/2024.05.03.592223 ( 2024 ). 21. ↵ Hayes , T. et al. Simulating 500 million years of evolution with a language model . Science, eads0018 ( 2025 ). 22. ↵ Nguyen , E. et al. Sequence modeling and design from molecular to genome scale with Evo . Science 386 , eado9336 ( 2024 ). OpenUrl CrossRef PubMed 23. ↵ Zvyagin , M. et al. GenSLMs: Genome-scale language models reveal SARS-CoV-2 evolutionary dynamics . The International Journal of High Performance Computing Applications 37 , 683 – 705 ( 2023 ). OpenUrl CrossRef 24. ↵ Buller , A. R. et al. Directed evolution of the tryptophan synthase β-subunit for stand-alone function recapitulates allosteric activation . Proceedings of the National Academy of Sciences 112 , 14599 – 14604 ( 2015 ). OpenUrl Abstract / FREE Full Text 25. ↵ Romney , D. K. , Murciano-Calles , J. , Wehrmüller , J. E. & Arnold , F. H. Unlocking reactivity of TrpB: a general biocatalytic platform for synthesis of tryptophan analogues . Journal of the American Chemical Society 139 , 10769 – 10776 ( 2017 ). OpenUrl CrossRef PubMed 26. ↵ Watkins-Dulaney , E. , Straathof , S. & Arnold , F. Tryptophan synthase: biocatalyst extraordinaire . ChemBioChem 22 , 5 – 16 ( 2021 ). OpenUrl CrossRef PubMed 27. ↵ Li , H. et al. Total Synthesis of Enlicitide Decanoate . Journal of the American Chemical Society 147 , 11036 – 11048 ( 2025 ). OpenUrl PubMed 28. ↵ Merck & Co., Inc . Merck Announces Positive Topline Results From the First Two Phase 3 CORALreef Trials Evaluating Enlicitide Decanoate for the Treatment of Adults With Hyperlipidemia Accessed: 2025-08-16 . 2025 . https://www.merck.com/news/merck-announces-positive-topline-results-from-the-first-two-phase-3-coralreef-trials-evaluating-enlicitide-decanoate-for-the-treatment-of-adults-with-hyperlipidemia/ . 29. ↵ Pickett , B. E. et al. ViPR: an open bioinformatics database and analysis resource for virology research . Nucleic Acids Research 40 , D593 – D598 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 30. ↵ Joshi , M. et al. Spanbert: Improving pre-training by representing and predicting spans . Transactions of the Association for Computational Linguistics 8 , 64 – 77 ( 2020 ). OpenUrl CrossRef 31. ↵ Johnson , S. R. et al. Computational scoring and experimental evaluation of enzymes generated by neural networks . Nature Biotechnology 43 , 396 – 405 ( 2025 ). OpenUrl CrossRef PubMed 32. ↵ Lin , Z. et al. Evolutionary-scale prediction of atomic-level protein structure with a language model . Science 379 , 1123 – 1130 ( 2023 ). OpenUrl CrossRef PubMed 33. ↵ Ahmed , S. A. , McPhie , P. & Miles , E. W. A Thermally Induced Reversible Conformational Transition of the Tryptophan Synthase Subunit Probed by the Spectroscopic Properties of Pyridoxal Phosphate and by Enzymatic Activity . Journal of Biological Chemistry 271 , 8612 – 8617 ( 1996 ). OpenUrl Abstract / FREE Full Text 34. ↵ Tang , S. , Tao , J. & Li , Y. Challenges and solutions for the downstream purification of therapeutic proteins . Antibody therapeutics 7 , 1 – 12 ( 2024 ). OpenUrl PubMed 35. ↵ Cao , E. , Chen , Y. , Cui , Z. & Foster , P. R. Effect of freezing and thawing rates on denaturation of proteins in aqueous solutions . Biotechnology and bioengineering 82 , 684 – 690 ( 2003 ). OpenUrl CrossRef PubMed Web of Science 36. ↵ Copley , S. D. Shining a light on enzyme promiscuity . Current Opinion in Structural Biology 47 , 167 – 175 ( 2017 ). OpenUrl CrossRef PubMed 37. ↵ Gardner , J. M. , Biler , M. , Risso , V. A. , Sanchez-Ruiz , J. M. & Kamerlin , S. C. Manipulating conformational dynamics to repurpose ancient proteins for modern catalytic functions . ACS Catalysis 10 , 4863 – 4870 ( 2020 ). OpenUrl 38. ↵ Scheele , R. A. et al. Ultrahigh throughput evolution of tryptophan synthase in droplets via an aptamer sensor . ACS Catalysis 14 , 6259 – 6271 ( 2024 ). OpenUrl PubMed 39. Goss , R. J. & Newill , P. L. A convenient enzymatic synthesis of L-halotryptophans . Chemical Communications , 4924 – 4925 ( 2006 ). 40. ↵ Winn , M. , Roy , A. D. , Grüschow , S. , Parameswaran , R. S. & Goss , R. J. A convenient one-step synthesis of L-aminotryptophans and improved synthesis of 5-fluorotryptophan . Bioorganic & Medicinal Chemistry Letters 18 , 4508 – 4510 ( 2008 ). OpenUrl PubMed 41. ↵ Smith , D. R. et al. The first one-pot synthesis of L-7-iodotryptophan from 7-iodoindole and serine, and an improved synthesis of other L-7-halotryptophans . Organic Letters 16 , 2622 – 2625 ( 2014 ). OpenUrl CrossRef PubMed 42. ↵ Almhjell , P. J. et al. The β-subunit of tryptophan synthase is a latent tyrosine synthase . Nature Chemical Biology 20 , 1086 – 1093 ( 2024 ). OpenUrl PubMed 43. ↵ Herger , M. et al. Synthesis of β-branched tryptophan analogues using an engineered subunit of tryptophan synthase . Journal of the American Chemical Society 138 , 8388 – 8391 ( 2016 ). OpenUrl PubMed 44. ↵ Murciano-Calles , J. , Romney , D. K. , Brinkmann-Chen , S. , Buller , A. R. & Arnold , F. H. A panel of TrpB biocatalysts derived from tryptophan synthase through the transfer of mutations that mimic allosteric activation . Angewandte Chemie International Edition 55 , 11577 – 11581 ( 2016 ). OpenUrl PubMed 45. ↵ Boville , C. E. , Romney , D. K. , Almhjell , P. J. , Sieben , M. & Arnold , F. H. Improved synthesis of 4-cyanotryptophan and other tryptophan analogues in aqueous solvent using variants of TrpB from Thermotoga maritima . The Journal of Organic Chemistry 83 , 7447 – 7452 ( 2018 ). OpenUrl PubMed 46. ↵ Rix , G. et al. Scalable continuous evolution for the generation of diverse enzyme variants encompassing promiscuous activities . Nature Communications 11 , 5644 ( 2020 ). OpenUrl PubMed 47. ↵ Wen , S. , Zheng , W. , Bornscheuer , U. T. & Wu , S. Generative artificial intelligence for enzyme design: Recent advances in models and applications . Current Opinion in Green and Sustainable Chemistry 52 , 101010 ( 2025 ). OpenUrl 48. ↵ Jurich , C. , Shao , Q. , Ran , X. & Yang , Z. J. Physics-based modeling in the new era of enzyme engineering . Nature Computational Science 5 , 279 – 291 ( 2025 ). OpenUrl PubMed 49. ↵ Hou , K. et al. De novo design of porphyrin-containing proteins as efficient and stereoselective catalysts . Science 388 , 665 – 670 ( 2025 ). OpenUrl CrossRef PubMed 50. ↵ Lauko , A. et al. Computational design of serine hydrolases . Science 388 , eadu2454 ( 2025 ). OpenUrl CrossRef PubMed 51. ↵ Amitai , G. , Gupta , R. D. & Tawfik , D. S. Latent evolutionary potentials under the neutral mutational drift of an enzyme . HFSP journal 1 , 67 ( 2007 ). OpenUrl PubMed 52. Turner , N.J. Directed evolution drives the next generation of biocatalysts . Nature Chemical Biology 5 , 567 – 573 ( 2009 ). OpenUrl PubMed 53. ↵ Bloom , J. D. , Romero , P. A. , Lu , Z. & Arnold , F. H. Neutral genetic drift can alter promiscuous protein functions, potentially aiding functional evolution . Biology Direct 2 , 17 ( 2007 ). OpenUrl PubMed 54. ↵ Rives , A. et al. Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences . Proceedings of the National Academy of Sciences 118 , e2016239118 ( 2021 ). OpenUrl Abstract / FREE Full Text 55. ↵ Alley , E. C. , Khimulya , G. , Biswas , S. , AlQuraishi , M. & Church , G. M. Unified rational protein engineering with sequence-based deep representation learning . Nature Methods 16 , 1315 – 1322 ( 2019 ). OpenUrl PubMed 56. ↵ Long , Y. et al. LevSeq: Rapid generation of sequence-function data for directed evolution and machine learning . ACS Synthetic Biology 14 , 230 – 238 ( 2024 ). OpenUrl PubMed 57. ↵ Rohl , C. A. , Strauss , C. E. , Misura , K. M. & Baker , D. in Methods in Enzymology 66 – 93 ( Elsevier , 2004 ). View the discussion thread. Back to top Previous Next Posted August 30, 2025. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Sequence-Based Generative AI-Guided Design of Versatile Tryptophan Synthases Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Sequence-Based Generative AI-Guided Design of Versatile Tryptophan Synthases Théophile Lambert , Amin Tavakoli , Gautham Dharuman , Jason Yang , Vignesh Bhethanabotla , Sukhvinder Kaur , Matthew Hill , Arvind Ramanathan , Anima Anandkumar , Frances H. Arnold bioRxiv 2025.08.30.673177; doi: https://doi.org/10.1101/2025.08.30.673177 Share This Article: Copy Citation Tools Sequence-Based Generative AI-Guided Design of Versatile Tryptophan Synthases Théophile Lambert , Amin Tavakoli , Gautham Dharuman , Jason Yang , Vignesh Bhethanabotla , Sukhvinder Kaur , Matthew Hill , Arvind Ramanathan , Anima Anandkumar , Frances H. Arnold bioRxiv 2025.08.30.673177; doi: https://doi.org/10.1101/2025.08.30.673177 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Biochemistry Subject Areas All Articles Animal Behavior and Cognition (7618) Biochemistry (17636) Bioengineering (13860) Bioinformatics (41847) Biophysics (21401) Cancer Biology (18536) Cell Biology (25424) Clinical Trials (138) Developmental Biology (13353) Ecology (19860) Epidemiology (2067) Evolutionary Biology (24287) Genetics (15583) Genomics (22463) Immunology (17701) Microbiology (40300) Molecular Biology (17141) Neuroscience (88434) Paleontology (666) Pathology (2825) Pharmacology and Toxicology (4813) Physiology (7633) Plant Biology (15107) Scientific Communication and Education (2042) Synthetic Biology (4285) Systems Biology (9808) Zoology (2268)