Transaminase engineering based on machine learning and rational design and applications thereof

Abstract

This invention relates to methods for modifying transaminases to improve their catalytic activity and substrate specificity, mutants, and their applications, belonging to the field of enzyme engineering technology. Starting from the parental transaminase CA032, whose amino acid sequence is shown in SEQ ID NO.1, this invention establishes a protein engineering modification model combining machine learning and rational design: constructing a preliminary mutant library through homology alignment; performing high-throughput virtual screening by coupling a machine learning (ML) fitness prediction model with physical stability filtering; and optimizing using multi-scale molecular dynamics (MD) simulations. The mutant CA032 M4 obtained through screening achieved greater than 99.9% conversion and enantioselectivity >99% ee at a high substrate loading of 100 g / L, providing an efficient biocatalytic scheme for the chiral synthesis of niraparib.

Description

Technical Field

[0001] This invention relates to the field of enzyme engineering technology, and in particular to a method for improving the catalytic activity and substrate specificity of transaminases through rational design strategies empowered by artificial intelligence / machine learning (AI / ML), the resulting mutants, and their application in the synthesis of chiral drug intermediates. Background Technology

[0002] Niraparib is a highly selective inhibitor of poly(ADP-ribose) polymerase (PARP) 1 and 2 that induces tumor cell apoptosis through a "synthetic lethality" mechanism. It is widely used clinically for maintenance therapy in advanced ovarian cancer, fallopian tube cancer, and primary peritoneal cancer. Due to its remarkable clinical efficacy, the efficient synthesis of niraparib and its key intermediates has become a focus of attention in the pharmaceutical industry.

[0003] Chiral piperidinones are key intermediates in the construction of the niraparib core framework. The core step in their synthesis involves converting the precursor sodium isopropyl ester substrate (CY006M1) into (S)-type chiral piperidinone (CY006-1) via chiral transamination. Currently, industrial synthesis primarily employs traditional chemical methods, which typically rely on noble metal catalysts (such as palladium and rhodium complexes) for asymmetric hydrogenation. This method is not only expensive but also involves stringent reaction conditions (such as high temperature and pressure), and faces environmental challenges such as heavy metal residues and wastewater treatment, making it difficult to meet the requirements of green chemistry and sustainable development.

[0004] In contrast, biocatalysis based on ω-transaminases (ω-TAs) has become the preferred alternative to chemical methods due to its extremely high chemo, regio, and enantioselectivity, as well as its mild reaction conditions and environmental friendliness. However, in practical applications, the synthesis of niraparib intermediates catalyzed by transaminases still faces the following key technical challenges:

[0005] 1. The conflict between activity and chiral selectivity:

[0006] In screening natural enzyme sources, researchers often face the dilemma of "high activity but incorrect chirality" or "correct chirality but low activity." For example, the transaminase CA048 obtained from the enzyme library constructed by our institution has high catalytic activity, but the chiral product it produces has the opposite conformation to the target intermediate. On the other hand, the transaminase CA032, which has the correct (S)-chiral conformation, has a narrow and compact active site, which has a natural advantage for catalyzing small natural substrates, but poor catalytic activity for the non-natural, large-volume substrates involved in this study.

[0007] 2. Poor compatibility with non-natural, large-volume substrates:

[0008] The niraparib intermediate, sodium isopropyl 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate, possesses a large skeletal structure. The active pocket space of the original enzyme CA032 cannot effectively accommodate this substrate, resulting in the Schiff base intermediate formed by the substrate and coenzyme PLP not being in an effective catalytic orientation. Furthermore, the substrate entry channel of the enzyme molecule exhibits significant geometrical constraints, generating severe mass transfer resistance and leading to extremely low catalytic efficiency, making it difficult to meet the loading requirements of industrial production.

[0009] 3. Limitations of rational design and evolutionary strategies:

[0010] Traditional directed evolution methods rely on random mutations and large-scale experimental screening. When faced with multi-site mutations with complex synergistic effects, they often fall into the bottleneck of "sequence space explosion," resulting in high experimental costs and low success rates. Existing rational design methods, if they rely solely on static structural analysis, often ignore the impact of protein dynamic flexibility on catalytic function, leading to insufficient stability of designed mutants in actual industrial systems (such as those containing cosolvents, high reaction temperatures, and long reaction times).

[0011] Therefore, how to utilize advanced AI / ML prediction technology coupled with multi-scale molecular dynamics simulation to precisely modify transaminases through "active pocket remodeling," "tunnel widening," and "conformation rigidification," and develop a transaminase mutant that can possess both high activity and chiral selectivity under high substrate loading, is a key issue that urgently needs to be addressed in the field of chiral drug intermediate biosynthesis. Summary of the Invention

[0012] This invention addresses the problems of low catalytic activity, poor pocket fit, and insufficient stability under high-load industrial conditions associated with the existing transaminase CA032 on the non-natural, bulky substrate 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate sodium salt (a key intermediate precursor for niraparib). It provides a series of transaminase mutants, their encoding genes, expression vectors, host cells, and their applications in the synthesis of chiral drug intermediates. Furthermore, this invention provides a protein engineering strategy that couples machine learning (ML) with multi-scale molecular dynamics (MD) simulations.

[0013] To achieve the above objectives, the present invention adopts the following technical solution:

[0014] A transaminase mutant, wherein the transaminase mutant is a transaminase CA032 with an amino acid sequence as shown in SEQ ID NO: 1, comprising amino acid substitutions at at least two sites from positions 8, 60, 61, 62, 65, 81, 94, 122, 124, 136, 169, 199, 209, 215, 217, 223, 269, 273, and 321. These amino acid substitutions are achieved through precise active pocket remodeling, substrate orientation regulation, and rigidification design, enabling the mutant to effectively accommodate large substrate volumes while maintaining the required catalytic geometry. The amino acid substitutions are selected from at least one of the following: Serine S at position 8 is mutated to threonine T, cysteine ​​C, or proline P; Tyrosine Y at position 60 is mutated to tryptophan W, isoleucine I, or phenylalanine F; Leucine L at position 61 is mutated to tyrosine Y, methionine M, or alanine A; Histidine H at position 62 is mutated to threonine T; Phenylalanine F at position 65 is mutated to alanine A or glycine G; Aspartic acid D at position 81 is mutated to glycine G or alanine A; Methionine M at position 94 is mutated to isoleucine I, leucine L, or threonine T; Phenylalanine F at position 122 is mutated to isoleucine I; Serine S at position 124 is mutated to isoleucine I, aspartic acid D, or glycine G; Phenylalanine F at position 136 is mutated to tryptophan W; Alanine A at position 169 is mutated to leucine L, valine V, or phenylalanine F; ... The following mutations were performed: valine (V) at position 209 was replaced by isoleucine (I); alanine (A) at position 209 was replaced by leucine (L); glycine (G) at position 215 was replaced by phenylalanine (F), lysine (K), or arginine (R); glycine (G) at position 217 was replaced by asparagine (N); serine (S) at position 223 was replaced by tryptophan (W), proline (P), or histidine (H); leucine (L) at position 269 was replaced by proline (P); leucine (L) at position 273 was replaced by phenylalanine (F), asparagine (N), or tyrosine (Y); and serine (S) at position 321 was replaced by glutamine (Q), proline (P), or tryptophan (W).

[0015] Furthermore, the mutant includes mutants consisting of any combination of 2, 3, 4 or more sites in (1)-(19) above;

[0016] Preferably, the mutant CA032 M2 contains a combination of mutations of M94I and F136W.

[0017] More preferably, the mutant CA032 M3 contains a combination of mutations of S8P, M94I, F136W, G217N, A209L, and S321P.

[0018] More preferably, the mutant CA032 M4 includes S8P, Y60W, L61Y, H62T, F65A, D81A, M94I, F122I, S124G, F136W, A169V, V199I, A209L, G215R, G217N, S223H, L269P, L273Y, and S321P.

[0019] The present invention also provides a gene encoding the above-mentioned transaminase mutant, wherein the nucleic acid sequence of the gene is derived from the CA032 gene sequence shown in SEQ ID NO.2, and a codon mutation is introduced at the corresponding site corresponding to the amino acid sequence shown in SEQ ID NO.1.

[0020] Furthermore, the present invention provides a recombinant expression vector containing the gene and a recombinant host cell, wherein the expression vector is a prokaryotic expression vector.

[0021] Preferably, the host cell is Escherichia coli BL21 (DE3).

[0022] The present invention also provides the use of the transaminase mutant, the gene, the expression vector, or the host cell in the asymmetric transamination reaction of the non-natural substrate 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate sodium isopropyl ester to generate the antitumor drug niraparib, which is a key chiral intermediate (S)-4-(4-bromophenyl)piperidin-2-one.

[0023] The present invention also provides a method for producing (S)-4-(4-bromophenyl)piperidin-2-one, comprising the following steps: adding the transaminase mutant to a reaction system containing sodium isopropyl 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate, an amino acid donor, and a cosolvent, wherein the pH is 9.0-10.0.

[0024] Preferably, the pH of the reaction system is 10.0.

[0025] Preferably, the concentration of the substrate is 50-150 g / L, and more preferably 100 g / L.

[0026] Preferably, the co-solvent is dimethyl sulfoxide (DMSO).

[0027] Preferably, the amino donor is isopropylamine, with an equivalent of 1.5-2.5 eq, more preferably 2.0 eq.

[0028] This invention also provides a method for modifying transaminases to improve their catalytic activity on non-natural, large-volume substrates, comprising the following steps: Step S1, Mutant library construction and data acquisition: Constructing a preliminary mutant library through homology alignment and alanine scanning, and collecting relative activity data of single mutants on target substrates; Step S2, Machine learning model training and prediction: Extracting the physicochemical and evolutionary conservation characteristics of residues, training a fitness prediction model using random forest (RF) and gradient boosting decision tree (XGBoost) algorithms, and performing high-throughput virtual screening on tens of thousands of multi-point mutation combinations; Step S3, Computational mechanism analysis and filtering: Analyzing the active pocket volume, near-attack conformation (NAC) frequency, and B-factor through molecular dynamics (MD) simulation, filtering the high-scoring sequences predicted by the model for stability and geometric fit, and determining the final mutation sites.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] 1. Catalytic performance is significantly improved.

[0031] The mutant CA032M4 obtained in this invention overcomes the steric hindrance of the parental enzyme to the non-natural, bulky substrate sodium 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate, increasing the active pocket volume from 850 Å. 3 Extended to 1120 Å 3 The catalytic efficiency is several orders of magnitude higher than that of the parent. It has high industrial application value. Under a high substrate loading of 100 g / L, the mutant conversion rate is greater than 99.9%, and the enantioselectivity of the product is extremely high (ee > 99%), which meets the requirements for industrial green synthesis of complex drug intermediates such as niraparib.

[0032] 2. Enhanced stability

[0033] By rigidifying the Loop region (e.g., L269P), the mutant exhibits excellent conformational stability in industrial systems containing organic cosolvents, at high reaction temperatures, and with long reaction times.

[0034] 3. The transformation strategy is efficient and precise.

[0035] This study demonstrates that, under “sparsely labeled” data conditions, coupling AI / ML prediction with physical mechanism filtering can significantly reduce the scale of experimental screening, providing a new and universal paradigm for protein engineering. Attached Figure Description

[0036] Figure 1 : Schematic diagram of the reaction.

[0037] Figure 2 Technology roadmap.

[0038] Figure 3 High-level structural comparison diagram of CA032 and CA084.

[0039] Figure 4 The relative activity and selectivity of the substrate at the differential sites of CA032 and CA084.

[0040] Figure 5 Heatmap analysis of the saturation mutation binding free energy predicted from 5Å residues in the active pocket.

[0041] Figure 6 : A diagram of the base material combined with the pocket.

[0042] Figure 7 RMSF analysis chart for CA032.

[0043] Figure 8 RMSF comparison diagram of CA032 wild type and CA032 M4.

[0044] Figure 9 HPLC chromatogram of the system after CA032 catalysis of the substrate.

[0045] Figure 10 HPLC chromatogram of the system after CA032 M4 catalyst substrate.

[0046] Figure 11 : Schematic diagram of H-NMR results, A: 1H NMR spectrum of CY006-1 product; B: 1H NMR spectrum of product after CA032M4 catalysis. Detailed Implementation

[0047] The technical solution of the present invention will be further described below through specific embodiments, but the scope of protection of the present invention is not limited thereto.

[0048] General Implementation Example:

[0049] Experimental Materials and General Methods

[0050] 1. Strains and plasmids:

[0051] Using E. coli BL21 (DE3) as the expression host, pET-28a(+) containing the CA032 gene (sequence shown in SEQ ID NO. 1) was used as the mutation template.

[0052] 2. Methods for constructing mutant libraries:

[0053] Site-directed mutagenesis was used to perform PCR amplification using the parental plasmid as a template. The PCR product was digested with DpnI and transformed into E. coli JM109 competent cells. Plasmids were extracted from single clones that were verified by sequencing and transformed into E. coli BL21(DE3) for expression. The specific procedures are as follows:

[0054] (1) Mutation primer design: Based on the amino acid sequence (SEQ ID NO.1) of CA032 and the target mutation site, site-directed mutation primers were designed using SnapGene software. The primers were synthesized by Genewiz Biotechnology Co., Ltd. (Suzhou, China).

[0055] (2) PCR amplification: The Fast site-directed mutagenesis kit (purchased from TransGen, Beijing, China) was used to perform PCR reaction with pET-28a(+) plasmid containing CA032 gene as template. The PCR reaction system (total volume 50 μL) was as follows: 10× DNA Polymerase Buffer 25 μL, 10 mM dNTP mixture (dATP, dCTP, dGTP and dTTP 2.5 mM each) 2 μL, forward primer and reverse primer with a concentration of 50 μM each 1 μL, template DNA 1 μL, DNA Polymerase 2 μL, ddH2O 18 μL.

[0056] (3) PCR reaction conditions: pre-denaturation at 95℃ for 5 min, denaturation at 95℃ for 30 s, annealing at 58℃ for 30 s, extension at 72℃ for 6 min, for a total of 30 cycles, and finally extension at 72℃ for 10 min.

[0057] (4) Template digestion and transformation: The PCR product was digested with enzymes at 37°C for 1 hour to completely consume the parental plasmid template; 5 μL of PCR product was added to 50 μL of E. coli JM 109 competent cell suspension in an ice bath and placed on ice for 30 min. The transformation product was heat-shocked at 42°C for 120 s and then quickly placed on ice to cool for 2 min. 500 μL of LB liquid medium was added to the tube and cultured at 37°C and 200 r / min for 60 min. 100 μL of the above bacterial solution was evenly spread on LB solid medium containing a final concentration of 50 μg / mL kanamycin. After the bacterial solution was completely absorbed by the medium, it was incubated upside down at 37°C for 12 h.

[0058] (5) Positive clone verification and plasmid extraction: single colonies on LB solid medium were picked and inoculated into LB liquid medium containing 50 μg / mL kanamycin. The culture was carried out at 37℃ and 200 r / min for 12 h. The plasmid was extracted using a plasmid mini-prep kit (purchased from TransGen, Beijing, China) and sent to Genewise Biotech Ltd. (Suzhou, China) for sequencing verification. The mutation site was confirmed to be correctly introduced, and a recombinant plasmid containing the target mutant gene was obtained.

[0059] (6) Construction of recombinant strains: Take 5 μL of correctly sequenced recombinant plasmid and add it to 50 μL of E. coli BL21(DE3) competent cell suspension in ice bath. Incubate on ice for 30 min, heat shock at 42℃ for 120 s, cool rapidly in ice bath for 2 min, add 500 μL of LB liquid medium, and incubate at 37℃ and 200 r / min for 60 min. Take 100 μL of bacterial solution and spread it on LB solid medium containing 50 μg / mL kanamycin. Incubate upside down at 37℃ for 12 h. After picking a single colony to verify that it is a positive clone, a recombinant genetically engineered strain that can express transaminase mutant is obtained.

[0060] 3. Protein expression and preparation of whole-cell enzyme powder:

[0061] The transformed strain was inoculated into LB medium containing 50 μg / mL kanamycin and cultured at 37°C until OD500. 600 The concentration was 0.6-0.8, and 0.5 mM IPTG was added. Expression was induced at 20℃ for 16 hours. Cells were collected by centrifugation, lyophilized, and used as whole-cell enzyme powder for later use.

[0062] 4. Activity detection:

[0063] The reaction system (1 mL) consisted of 100 mM Tris-HCl buffer, pH adjusted to approximately 10.0 with isopropylamine, 10 mM sodium isopropyl 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate, 0.1 mM PLP, and an appropriate amount of whole-cell lyophilized enzyme powder. The substrate conversion rate and product ee value were determined using high-performance liquid chromatography (HPLC). The specific detection methods are as follows:

[0064] (1) Detection instrument: High performance liquid chromatograph

[0065] (2) Column type: XBridge®Shield RP18 4.6*250 mm 5 μm;

[0066] (3) Mobile phase: Phase A is methanol, and Phase B is pure water;

[0067] (4) Detection parameters: flow rate 1 mL / min, detection time 20 min, column temperature 40℃, detection wavelength 210 / 220 nm, injection volume 15 μL;

[0068] (5) Gradient elution procedure:

[0069] A% 30.0 80.0 30.0 30.0 B% 70.0 20.0 70.0 70.0

[0070] (6) Detection and calculation: After the reaction is completed, take 200 μL of the reaction solution, add 1 mL of methanol to terminate the reaction, centrifuge at 12000 r / min for 10 min, take the supernatant and filter it through a 0.22 μm filter membrane for HPLC detection.

[0071] Example 1: Virtual Screening of Mutants Based on AI / ML

[0072] 1. Preliminary data acquisition:

[0073] Site-directed mutagenesis was performed on 30 differential residues between CA032 and its homolog CA048 to investigate the relative catalytic activity and chiral selectivity of 30 single mutants to the substrate sodium isopropyl 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate.

[0074] 2. Feature Engineering and Modeling:

[0075] Physicochemical features of residues (including side chain volume, hydrophobicity, etc.) and evolutionary features of PSSM based on homologous sequence alignment were extracted. An ensemble learning prediction model was constructed using Random Forest and Gradient Boosting Decision Tree (XGBoost) algorithms.

[0076] 3. Virtual screening and combination design:

[0077] Within a 5 Å range of the active site, the binding free energy of saturation mutations was predicted, generating potential two-point and multi-point mutant combinations. Model predictions showed that the combinatorial mutant M94I / F136W (named CA032M2) had the highest fitness score and was predicted to have the highest fold increase in substrate activity. This was confirmed by subsequent experiments.

[0078] Example 2: Active Pocket Remodeling and Geometric Fit Analysis

[0079] 1. Simulation Analysis

[0080] Molecular docking simulations were performed between the CA032 crystal structure and the bulky substrate sodium isopropyl 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate. The results showed that the CA032 enzyme's active pocket space is limited and cannot accommodate the large substrate backbone structure. Specifically, the side chains of key residues Y60, L61, F122, S124, and L273 located in the pocket wall and entrance exhibit severe van der Waals force conflicts with the benzene ring and ester backbone of the substrate. This makes it difficult for the substrate molecule to penetrate deep into the catalytic center and form a stable Schiff base intermediate with the coenzyme PLP, which may be the fundamental structural cause of the low activity of the parent enzyme.

[0081] 2. Implementation of the renovation:

[0082] To address the aforementioned bottleneck sites, this invention employs precise steric hindrance reduction modifications (such as F122I, S124G, and L61A). Calculations and verifications show that these site mutations significantly expand the effective volume of the active pocket, allowing the substrate to enter the pocket in a more optimized conformation, and significantly shortening the critical catalytic distance (between the amino donor and the carbonyl carbon atom).

[0083] 3. Computational dynamics verification:

[0084] 100 ns molecular dynamics (MD) simulations showed that the near-attack conformation frequency of the Schiff base intermediate formed by the substrate and coenzyme PLP in the CA032 M4 mutant was significantly increased from 12% to 68%, demonstrating that pocket remodeling effectively improved catalytic efficiency.

[0085] Example 3: Substrate orientation regulation and chiral selective flipping based on active site remodeling

[0086] 1. Analysis of conformational regulation mechanisms:

[0087] Molecular simulations and structural biology analyses revealed that amino acid residue 136 is located at a critical position in the active site of the CA032 transaminase, responsible for guiding the geometric orientation of substrate molecules when they enter the binding pocket. In the parent enzyme (WT), phenylalanine (F136) is present at position 136. Its small benzene ring side chain provides some binding space for the bulky substrate, sodium isopropyl 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate, but lacks sufficient spatial constraint to induce a specific catalytic conformation. This results in poor substrate binding orientation within the pocket, and may even lead to multiple competing orientations, limiting the overall selectivity of the enzyme.

[0088] 2. Mutant design and the introduction of tryptophan (W):

[0089] This invention replaces phenylalanine at position 136 with tryptophan (F136W) through site-directed mutagenesis. Unique side-chain effect: Tryptophan possesses a larger indole group with a higher electron cloud density. The indole ring not only enhances aromatic interactions within the pocket (“pi-pi” stacking), but more importantly, due to its unique bicyclic rigid structure, it generates a significant spatial remodeling effect within the active pocket. Substrate conformational inversion: The indole group on the mutated tryptophan side chain occupies a specific subregion of the original pocket, producing a synergistic effect of “spatial compression” and “orientation guidance.” This effect forcibly causes an approximately 180° conformational inversion of the substrate's large aromatic skeleton upon binding, bringing its carbonyl center closer to the coenzyme PLP with a novel geometric orientation, thereby achieving precise control of catalytic chiral selectivity.

[0090] 3. Experimental Results and Selective Validation:

[0091] Experimental data show that the introduction of the F136W mutant significantly altered the substrate selectivity of CA032, successfully achieving a high enrichment of product configuration.

[0092] Example 4: Conformational Flexible Control and Local Rigidification Design Based on RMSF Analysis

[0093] 1. Flexible regional positioning and bottleneck analysis:

[0094] A 100 ns molecular dynamics (MD) simulation was performed on wild-type (WT) CA032 transaminase, and the root mean square fluctuation (RMSF) values ​​of amino acid residues were calculated. The results showed that the loop region composed of residues 265-275 of CA032 exhibited significant fluctuation peaks, with an RMSF value higher than the overall protein average. This indicates its extremely high conformational flexibility in an aqueous environment. Because this loop region is close to the edge of the active pocket, its violent random fluctuations make it difficult for the active pocket to maintain a stable geometry during catalysis, thus interfering with the precise capture and binding of the non-natural, bulky substrate, sodium isopropyl 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate.

[0095] 2. Rigidification Strategy – Introduction of Proline:

[0096] To address the aforementioned highly volatile region, this invention selected leucine residue 269 (L269), located at the core node of the loop region, as the modification site. An L269P mutation was introduced (mutating leucine to proline). Its side chain is linked to the nitrogen atom of the peptide bond, forming a rigid constraint. This structure significantly restricts the rotational freedom of the protein backbone, thus acting as a "fixed anchor" at the molecular level and inhibiting excessive flipping of the entire loop region.

[0097] 3. Dynamic stability verification:

[0098] Further MD simulations showed that the RMSF value of the CA032 M4 mutant in the 265-275 region was significantly lower than that of the wild type, demonstrating effective rigidification of the local region. Experimental results: Kinetic experiments showed that this local conformational stabilization effectively maintained the geometric integrity of the active pocket during the reaction process. Experimental verification showed that the CA032 M4 mutant exhibited superior structural stability in a system containing 10% DMSO. This demonstrates that by inhibiting the flexibility of the key loop region, the industrial catalytic performance of transaminases on complex substrates can be effectively improved.

[0099] Example 5: Industrial-grade performance verification under 100 g / L load

[0100] 1. Construction of the reaction system:

[0101] In a 100 mL reactor, add 100 g / L of substrate, 10% DMSO as a co-solvent, CA032M4 enzyme lyophilized powder (enzyme / substrate ratio of 50%), and isopropylamine as an amino donor (2.0 eq). The pH is automatically titrated to maintain at 10.0.

[0102] 2. Dynamic monitoring of conversion:

[0103] The conversion rate reached 72% after 24 hours of reaction.

[0104] After 48 hours of reaction, HPLC analysis showed complete substrate conversion (conversion rate >99.9%).

[0105] 3. Product characterization:

[0106] The product was extracted with ethyl acetate and purified by column chromatography to obtain a white powder. The structure of the product was characterized by 1H NMR spectroscopy, and its chemical shift and integrated area were completely consistent with the theoretical structural characteristics of the target product (S)-4-(4-bromophenyl)piperidin-2-one, confirming the successful synthesis. Chiral HPLC analysis showed an enantiomeric excess (ee) of 99.5%. These results demonstrate that CA032M4 fully meets the requirements for large-scale biocatalytic synthesis of niraparib intermediates.

Claims

1. A transaminase mutant, characterized in that, The transaminase mutant is a transaminase CA032 with an amino acid sequence as shown in SEQ ID NO.1, comprising amino acid substitutions at at least two sites among positions 8, 60, 61, 62, 65, 81, 94, 122, 124, 136, 169, 199, 209, 215, 217, 223, 269, 273, and 321. The amino acid substitutions include at least one of the following types: serine at position 8 is mutated to threonine, cysteine, or proline; tyrosine at position 60 is mutated to tryptophan, isoleucine, or phenylalanine; leucine at position 61 is mutated to tyrosine, methionine, or alanine; histidine at position 62 is mutated to threonine; phenylalanine at position 65 is mutated to alanine or glycine; aspartic acid at position 81 is mutated to glycine or alanine; and methionine at position 94 is mutated to isoleucine. Leucine, leucine, threonine; phenylalanine at position 122 is mutated to isoleucine; serine at position 124 is mutated to isoleucine, aspartic acid, glycine; phenylalanine at position 136 is mutated to tryptophan; alanine at position 169 is mutated to leucine, valine, phenylalanine; valine at position 199 is mutated to isoleucine; alanine at position 209 is mutated to leucine; glycine at position 215 is mutated to phenylalanine, lysine, arginine; glycine at position 217 is mutated to asparagine; serine at position 223 is mutated to tryptophan, proline, histidine; leucine at position 269 is mutated to proline; leucine at position 273 is mutated to phenylalanine, asparagine, tyrosine; serine at position 321 is mutated to glutamine, proline, tryptophan.

2. The transaminase mutant according to claim 1, characterized in that, The mutant CA032M4 contains a combination of mutations including S8P, Y60W, L61Y, H62T, F65A, D81A, M94I, F122I, S124G, F136W, A169V, V199I, A209L, G215R, G217N, S223H, L269P, L273Y, and S321P.

3. A gene encoding the transaminase mutant of claim 1 or 2, wherein the nucleic acid sequence of the gene is derived from the CA032 gene sequence shown in SEQ ID NO. 2, and a codon mutation is introduced at the corresponding site corresponding to the amino acid sequence shown in SEQ ID NO.

1.

4. The recombinant expression vector of the gene according to claim 3 and the recombinant host cell, wherein the expression vector is a prokaryotic expression vector.

5. The recombinant host cell according to claim 4, characterized in that, The host cell was Escherichia coli E. coli BL21 (DE3).

6. The use of the transaminase mutant of claim 1, the gene of claim 3, the recombinant expression vector of claim 4, or the recombinant host cell of claim 5 in a catalytic substrate.

7. The application according to claim 6, characterized in that: The substrate is sodium salt of isopropyl 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate, a non-natural, bulky substrate, and the product is (S)-4-(4-bromophenyl)piperidin-2-one.

8. A method for producing (S)-4-(4-bromophenyl)piperidin-2-one, characterized in that: The transaminase mutant of claim 1 or 2 or an expression system containing the mutant is added to a reaction system containing sodium isopropyl 3-(4-bromophenyl)-4-hydroxy-4-(sulfinyloxy)valerate, and an amino acid donor and a cosolvent are added to carry out the reaction.

9. A method for modifying transaminases to improve their catalytic activity on large-volume non-natural substrates, characterized in that, Includes the following steps: Step S1: Construct a preliminary mutant library containing key residues through homology comparison; Step S2: Couple machine learning fitness prediction model with physical stability filtering to perform high-throughput virtual screening; Step S3: Combine multi-scale molecular dynamics simulation and structural biology analysis to evaluate the active pocket, substrate channel and backbone rigidity of the screened mutants.

10. The method according to claim 9, characterized in that: In step S2, random forest and gradient boosting decision tree are used for ensemble learning prediction.

Images