TY - JOUR
T1 - Precision Thermostability Predictions
T2 - Leveraging Machine Learning for Examining Laccases and Their Associated Genes
AU - Tiwari, Ashutosh
AU - Krisnawati, Dyah Ika
AU - Widodo,
AU - Cheng, Tsai Mu
AU - Kuo, Tsung Rong
N1 - Publisher Copyright:
© 2024 by the authors.
PY - 2024/12
Y1 - 2024/12
N2 - Laccases, multi-copper oxidases, play pivotal roles in the oxidation of a variety of substrates, impacting numerous biological functions and industrial processes. However, their industrial adoption has been limited by challenges in thermostability. This study employed advanced computational models, including random forest (RF) regressors and convolutional neural networks (CNNs), to predict and enhance the thermostability of laccases. Initially, the RF model estimated melting temperatures with a training mean squared error (MSE) of 13.98, and while it demonstrated high training accuracy (93.01%), the test and validation MSEs of 48.81 and 58.42, respectively, indicated areas for model optimization. The CNN model further refined these predictions, achieving lower training and validation MSEs, thus demonstrating enhanced capability in discerning complex patterns within genomic sequences indicative of thermostability. The integration of these models not only improved prediction accuracy but also provided insights into the critical determinants of enzyme stability, thereby supporting their broader industrial application. Our findings underscore the potential of machine learning in advancing enzyme engineering, with implications for enhancing industrial enzyme stability.
AB - Laccases, multi-copper oxidases, play pivotal roles in the oxidation of a variety of substrates, impacting numerous biological functions and industrial processes. However, their industrial adoption has been limited by challenges in thermostability. This study employed advanced computational models, including random forest (RF) regressors and convolutional neural networks (CNNs), to predict and enhance the thermostability of laccases. Initially, the RF model estimated melting temperatures with a training mean squared error (MSE) of 13.98, and while it demonstrated high training accuracy (93.01%), the test and validation MSEs of 48.81 and 58.42, respectively, indicated areas for model optimization. The CNN model further refined these predictions, achieving lower training and validation MSEs, thus demonstrating enhanced capability in discerning complex patterns within genomic sequences indicative of thermostability. The integration of these models not only improved prediction accuracy but also provided insights into the critical determinants of enzyme stability, thereby supporting their broader industrial application. Our findings underscore the potential of machine learning in advancing enzyme engineering, with implications for enhancing industrial enzyme stability.
KW - computational genomics
KW - convolutional neural network
KW - enzyme engineering
KW - laccase thermostability
KW - machine learning
KW - random forest regression
KW - computational genomics
KW - convolutional neural network
KW - enzyme engineering
KW - laccase thermostability
KW - machine learning
KW - random forest regression
UR - http://www.scopus.com/inward/record.url?scp=85211784727&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85211784727&partnerID=8YFLogxK
U2 - 10.3390/ijms252313035
DO - 10.3390/ijms252313035
M3 - Article
C2 - 39684743
AN - SCOPUS:85211784727
SN - 1661-6596
VL - 25
JO - International journal of molecular sciences
JF - International journal of molecular sciences
IS - 23
M1 - 13035
ER -