Protein Engineering of Amine Transaminases and Methyltransferases using Machine Learning and High-Throughput Screening Tools - Inaugural Dissertation zur Erlangung des akademischen Grades eines Doktors der Naturwissenschaften (Dr. rer. nat.)

Aug 2, 2026

Upload by Ezra

This inaugural dissertation investigates how machine learning combined with high-throughput screening can improve protein engineering of amine transaminases and SAM-dependent methyltransferases. It reviews traditional directed evolution and (semi-)rational design, then details dataset and algorithm considerations for ML predictors, including transfer learning and key limitations of data-driven approaches. The work further develops and validates an information-rich training strategy and an assay workflow to enable efficient data generation, followed by experimental verification of predicted variants and enzyme performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages190

Reading Time7 h 59 min

AccessFree to download

Scope & Outline

Introduction

Traditional Protein Engineering in Biotechnology

Machine learning – An emerging new paradigm

Transferases

Results & Discussion

Development of an amine transaminase ML predictor (Article I)

Further investigation of the potential of the ML predictor (Article II)

Current limitations for data-driven protein engineering (Article III)

Development of a universal SAM-dependent methyltransferase assay (Article IV)

Conclusion

References

Protein Engineering of Amine Transaminases and Methyltransferases using Machine Learning and High-Throughput Screening Tools I n a u g u r a l d i s s e r t a t i o n zur Erlangung des akademischen Grades eines Doktors der Naturwissenschaften (Dr. rer. nat. ) der Mathematisch-Naturwissenschaftlichen Fakultät der Universität Greifswald vorgelegt von Marian Jendrik Menke Greifswald, Juni 2024 Dekan: Prof. Dr. Matthias Eschrig 1. Gutachter: Prof. Dr. Uwe T. Bornscheuer 2. Gutachter: Jun.-Prof. Dr. Stephan Hammer 3. Gutachter: Prof. Dr. Ioannis Pavlidis Tag der Promotion: 16.09.2024 Table of contents List of Abbreviations ............................................................................................................................... III Scope & Outline ....................................................................................................................................... V 1. Introduction ..................................................................................................................................... 1 1.1 Traditional Protein Engineering in Biotechnology ................................................................... 1 1.1.1 Directed Evolution ........................................................................................................... 2 1.1.2 (Semi-)Rational Design .................................................................................................... 3 1.2 Machine learning – An emerging new paradigm .................................................................... 6 1.2.1 Dataset Considerations.................................................................................................... 7 1.2.2 Algorithm Considerations ................................................................................................ 9 1.3 Transferases ........................................................................................................................... 11 1.3.1 Transaminases ............................................................................................................... 11 1.3.2 Structural basis of amine transaminases for perfect enantioselectivity ....................... 12 1.3.3 Enzymatic mechanism and reaction modes of amine transaminases........................... 13 1.3.4 Methyltransferases are promising biocatalysts ............................................................. 17 1.3.5 Industrial limitations of SAM-dependent NPMTs .......................................................... 18 1.3.6 Methyltransferase Assays .............................................................................................. 19 2. Results & Discussion ...................................................................................................................... 23 2.1 Development of an amine transaminase ML predictor (Article I) ......................................... 23 2.1.1 Creation of a small and diverse dataset with inverted stereoselectivity....................... 23 2.1.2 Creation of the ML predictor and prediction of improved variants .............................. 26 2.1.3 Transfer learning and limitations of the simple ML algorithm ...................................... 28 2.2 Further investigation of the potential of the ML predictor (Article II) .................................. 28 2.2.1 Creation of a small, but information-rich training dataset ............................................ 28 2.2.2 Building of the ML predictor & In silico sequence-ﬁtness space screening .................. 31 2.2.3 Experimental validation ................................................................................................. 31 2.3 Current limitations for data-driven protein engineering (Article III) ..................................... 33 2.4 Development of a universal SAM-dependent methyltransferase assay to facilitate data

FAQs

What problem does this dissertation address in enzyme engineering?

It focuses on engineering amine transaminases and SAM-dependent methyltransferases using machine learning and high-throughput screening to generate improved enzyme variants and reliable experimental outcomes.

How is machine learning applied to predict improved transaminase variants?

The dissertation builds an ML predictor using a small but diverse dataset with inverted stereoselectivity, then evaluates creation of the predictor, prediction of improved variants, and the role and limits of transfer learning and algorithm simplicity.

What role does the methyltransferase assay development play?

It addresses the need for efficient data generation by designing a universal SAM-dependent methyltransferase assay, evaluating assay principles stepwise with lysate-based validation and outlining considerations for further improvements.

Aug 2, 2026

Upload by Ezra