아스파라기나아제 효소 활성 향상을 위한 AI 기반 진화적 스케일 모델(ESM) 적용 연구
Enhancement of Asparaginase Activity via the AI-Based Evolutionary Scale Model (ESM)
- 주제(키워드) 도움말 L-asparaginase , ESM
- 발행기관 국립강릉원주대학교 일반대학원
- 지도교수 도움말 연영주
- 발행년도 2025
- 학위수여년월 2026. 2
- 학위명 석사
- 학과 및 전공 도움말 일반대학원 생명화학공학과
- 세부분야 해당없음
- 실제URI http://www.dcollection.net/handler/kangnung/000000012320
- UCI I804:42001-000000012320
- 본문언어 영어
초록/요약 도움말
Escherichia coli–derived L-asparaginase II (EcAII) is an enzyme encoded by the L-asparaginase II (ansB) gene that catalyzes the hydrolysis of L-asparagine into L-aspartate and ammonia, thereby reducing circulating asparagine levels and inhibiting the proliferation of acute lymphoblastic leukemia (ALL) cells. While normal cells are capable of synthesizing asparagine endogenously, ALL cells depend on exogenous asparagine; thus, depletion of this amino acid results in inhibition of protein synthesis, leading to growth arrest and cell death. Through this selective mechanism of action, EcAII precisely targets the metabolic vulnerability of leukemia cells. In this study, Evolutionary Scale Model 2 (ESM-2) was applied to enhance the enzymatic activity of EcAII. ESM-2 is a transformer-based language model pretrained on large-scale protein sequence datasets, enabling it to learn statistical correlations within protein sequences and long-range residue interactions. Without relying on structural information, ESM-2 can predict the relative fitness of amino acid substitutions based on sequence information and assess the functional impact of mutations through analyses based on changes in sequence probability. ESM-2–based mutation effect analysis was performed on the EcAII sequence, and residue-level scores were postprocessed using the ESM-based Residue Prioritization using Total and Positive Averages (ESM-TPA) approach to identify candidate residues with low evolutionary constraint and high potential for improvement. A variant library constructed based on the selected residues was subsequently evaluated through experimental screening, leading to the identification of variants exhibiting enhanced catalytic activity compared to the wild-type. Successive site-saturation mutagenesis was performed at the identified positions, and activity assessments revealed high-performance variants with improved activity relative to the wild-type. In conclusion, this study demonstrates that ESM-2–based mutation effect prediction with experimental screening is effective in engineering enzymatic activity. This approach provides an effective AI-based approach for accelerating large-scale protein variant exploration and functional protein design.
more목차 도움말
1. Introduction 15
2. Literature survey 19
2.1. L-asparaginase II 19
2.1.1. Structural and Mutational Insights 19
2.1.2. Anticancer Mechanisms and Clinical Implications 24
2.1.3 Applications in biomedical, food, and industrial fields 29
2.2. Protein Language Model (PLM) 33
2.2.1 Transformer-based model (Encoder-only model) 37
2.2.2 Transformer-based model (Decoder-only model) 42
2.2.3 Transformer-based model (Encoder-Decoder model) 44
2.3. Evolutionary Scale Modeling (ESM) 47
3. Materials & Methods 50
3.1. Materials 50
3.2. ESM-2–Based Mutation Site Selection 51
3.3. Gene cloning and Mutagenesis 53
3.4. Expression, and Purification of the recombinant EcAII 61
3.5. Asparaginase Activity Assay 63
4. Results & discussion 66
4.1. Mutation, Expression and Purification of EcAII 66
4.2. In vitro Activity Screening of ESM-Selected EcAII Mutants 78
4.3. In vitro Enzymatic Activity of Site-Saturation Variants 80
4.4. Enzyme Structural Analysis 82
5. Conclusion 89
6. References 91
