The development of synthetic biology relies on accurate predictions of enzyme kinetic parameters, such as turnover number, Michaelis constant, and catalytic efficiency. Traditional models have suffered from inaccuracy and overfitting, making it challenging to obtain reliable predictions. To address this, a new tool called CataPro was introduced, which uses deep learning and molecular fingerprints to predict enzyme kinetic parameters. CataPro has been shown to outperform previous models, demonstrating enhanced accuracy and generalization ability on unbiased datasets. The tool uses pre-trained models and molecular structure data of substrates to analyze amino acid sequences of enzymes, avoiding data leakage and improving predictions.

The implications of this advancement go beyond academic research, with potential applications in industry, such as improved biocatalyst development and optimized processes. Enzymes are crucial in industrial applications, including pharmaceuticals, biofuels, and food production. The introduction of CataPro is a significant step forward in enzyme kinetic parameter prediction, having the potential to revolutionize enzyme discovery and modification. With ongoing refinements, CataPro may play a crucial role in synthetic biology, enabling efficient and effective solutions that meet industrial demands while promoting environmental sustainability.

Read the Full Article

The paper introduces CataPro, a neural network-based framework for predicting enzyme kinetic parameters, such as kcat and Km. The framework uses a combination of language models and molecular fingerprints to encode enzyme and substrate information. The model is trained on an unbiased dataset of enzyme kinetic data and demonstrates improved performance compared to existing methods, such as DLKcat and UniKP. The paper also presents a strategy for predicting kcat/Km by incorporating a neural network-based correction term. The model is evaluated on a range of enzymes and substrates, and is shown to outperform existing methods in terms of metrics such as Pearson’s correlation coefficient and root-mean-squared error.

The paper then demonstrates the application of CataPro in enzyme discovery and engineering, using the decarboxylation of ferulic acid to produce 4-VG as an example. The model is used to select candidates for the oxidation of 4-VG, and five representative enzymes are identified. The paper also presents the use of CataPro in enzyme engineering, using the SsCSO enzyme to improve its activity and selectivity. The results demonstrate the effectiveness of CataPro in enzyme discovery and engineering, with the potential to accelerate the discovery of new enzymes and improve their activity and selectivity.

Read the Full Article

A team of bioengineers and synthetic biologists has developed a machine-learning guided platform to design and test thousands of new enzymes quickly and efficiently. The platform, described in a paper titled “Accelerated enzyme engineering by machine-learning guided cell-free expression,” uses machine learning to predict the behavior of enzymes and test their performance in various chemical reactions. This approach allows for rapid iteration and optimization of enzyme design, bypassing traditional methods that require manual modification of DNA and testing in living cells. The platform was used to synthesize a small-molecule pharmaceutical at 90% yield, a significant improvement over previous attempts. The researchers also demonstrated the ability to generate multiple specialized enzymes in parallel to produce eight additional therapeutics. The potential applications of this technology are vast, including the development of sustainable and efficient processes in industries such as pharmaceuticals, food production, and environmental remediation. While there are still challenges to overcome, including lack of high-quality data, the researchers believe that machine learning can revolutionize the field of enzyme engineering and accelerate the development of new technologies.

Read the Full Article

A team of bioengineers and synthetic biologists has developed a computational workflow that can design thousands of new enzymes, predict their behavior in the real world, and test their performance across multiple chemical reactions, all on a computer. This breakthrough could revolutionize the enzyme engineering field, which is crucial for many industries, including pharmaceuticals, energy, and the environment. Traditionally, scientists would use living cells to produce enzymes, a process that is time-consuming and costly. With this new approach, scientists can use machine learning to predict the performance of enzymes and test them quickly, potentially reducing the time and cost of enzyme development. The team’s proof-of-concept study demonstrated the success of their approach, synthesizing a small-molecule pharmaceutical at 90% yield, compared to an initial 10% yield. The potential applications of this technology are vast, including the development of new enzymes for sustainable energy, cleaner air and water, and more efficient pharmaceuticals. While there are still some challenges to overcome, this breakthrough has the potential to accelerate the discovery of new enzymes and improve the efficiency of many industries.

Read the Full Article

Scientists from Northwestern Engineering and Stanford University have made a breakthrough in the creation of amide bonds, a fundamental component of many natural and synthetic materials. Amide bonds are used in proteins, pharmaceuticals, and everyday products like agrochemicals, fragrances, and flavors. Researchers, led by Ashty Karim and Michael Jewett, have developed a platform to engineer enzymes responsible for forming amide bonds, which could revolutionize the field of green chemistry. The platform uses a novel high-throughput, cell-free, and machine-learning-guided approach to rapidly generate large datasets and predict the function of enzymes. The team used this platform to engineer 1,217 mutants of an amide synthetase enzyme, McbA, to form nine small molecule pharmaceuticals. This achievement demonstrates the versatility of McbA to catalyze many unique reactions and the ability to rapidly build specialized biocatalysts in parallel. The researchers believe that this work has the potential to transform the bioeconomy across various industries in energy, materials, and medicine. However, more research is needed to improve the approach and explore new artificial intelligence methods to create new-to-nature proteins.

Read the Full Article

A recent study has combined machine learning with cell-free gene expression systems to accelerate the optimization of enzymes for various chemical reactions. The framework, called DBTL (design-build-test-learn), enables the simultaneous evaluation of multiple enzyme variants, greatly increasing the speed and efficiency of enzyme engineering. The researchers used a dataset of over 1,200 enzyme variants and more than 10,900 reactions to train their machine learning models, which accurately predicted variants likely to excel at producing specific small molecules. In tests, ML-optimized mutants showed improvements ranging from 1.6 to 42-fold for the synthesis of nine different pharmaceuticals. The ML-guided framework has the potential to revolutionize the field of enzyme engineering, allowing for the rapid creation of specialized biocatalysts across various domains. This approach could particularly benefit pharmaceutical applications, where cost-effective and sustainable production methods are crucial. Overall, the integration of machine learning and cell-free gene expression has the potential to accelerate the discovery of novel enzymes and their applications in various industries.

Read the Full Article

Researchers have designed and synthesized a task-specific PEG-IL-based liquid-solid hybrid microreactor for enhanced continuous-flow biocatalysis. The microreactor is composed of Pickering droplets of PEG-IL and hydrophobic silica nanoparticles, which encapsulate Candida antarctica lipase B (CALB) enzyme. The PEG-ILs are synthesized by reacting methylimidazolium cations with hydrophobic silica nanoparticles and are used as the dispersed phase. The hybrid microreactor exhibits excellent stability, with no significant fluctuations in enzyme activity over a period of 320 hours.

The microreactor shows enhanced activity and thermal stability compared to a PEG-free IL-based catalyst. The PEG-ILs play a critical role in stabilizing the CALB enzyme, reducing conformational changes, and retaining its structure at high temperatures. The microreactor is applied to the kinetic resolution of various racemic alcohols, including a pharmaceutical alcohol intermediate, with high ee values (above 99%) maintained over a long period.

The microreactor’s morphology and structure are essentially unchanged after reaction, and its PEG-IL content remains the same. The researchers believe that the task-specific PEG-IL-based liquid-solid hybrid microreactor has great potential for real-world applications in biocatalysis and pharmaceutical manufacturing.

Read the Full Article