TY - JOUR
T1 - Improving accuracy scores of neural network driven QSAR models of mutagenicity
AU - Kalian, Alex
AU - Osborne, Olivia J
AU - Dorne, Jean-Lou C. M.
AU - Gott, David
AU - Potter, Claire
AU - Guo, Miao
AU - Hogstrand, Christer
N1 - Funding Information:
This work was supported by grants from the Biotechnology and Biological Sciences Research Council [grant number BB/T008709/1] and the Food Standards Agency [Agency Project FS900120], The views expressed in this article do not reflect the views of the European Food Safety Authority (EFSA) and/or are a reflection of the views of the authors only. This paper aims to contribute to the international network on Advancing the Pace of Chemical Risk Assessment (APCRA), to contribute to the use of New Approach Methodologies (NAMs) in chemical risk assessment and ultimately reduce animal testing.
Funding Information:
This work was supported by grants from the Biotechnology and Biological Sciences Research Council [grant number BB/T008709/1] and the Food Standards Agency [Agency Project FS900120]
Publisher Copyright:
© 2023 Elsevier B.V.
PY - 2023/7/18
Y1 - 2023/7/18
N2 - Multiple QSAR models of mutagenicity were created and compared, using knowledge graph approaches to train and test multi-layer perceptron classifiers, following dimensionality reduction from several thousand dimensions to hundreds of dimensions via principal component analysis. Such knowledge graphs were built in one case using molecular fingerprint based structural similarities, while in another case using molecular fragments found via application of the Girvan-Newman algorithm. A simple hybrid model was also explored. However, both competing QSAR models performed with comparable accuracies, with both sensitivity and specificity scores for each occurring within range of 70%. The predictions of both models were in agreement in an average of 71% of cases, meaning that each could offer a related yet notably different perspective of toxicological space; hence a simple hybrid model was trialed, which only output predictions agreed between both constituent models, which averaged at 78% accuracy.
AB - Multiple QSAR models of mutagenicity were created and compared, using knowledge graph approaches to train and test multi-layer perceptron classifiers, following dimensionality reduction from several thousand dimensions to hundreds of dimensions via principal component analysis. Such knowledge graphs were built in one case using molecular fingerprint based structural similarities, while in another case using molecular fragments found via application of the Girvan-Newman algorithm. A simple hybrid model was also explored. However, both competing QSAR models performed with comparable accuracies, with both sensitivity and specificity scores for each occurring within range of 70%. The predictions of both models were in agreement in an average of 71% of cases, meaning that each could offer a related yet notably different perspective of toxicological space; hence a simple hybrid model was trialed, which only output predictions agreed between both constituent models, which averaged at 78% accuracy.
KW - cheminformatics
KW - QSAR
KW - neural network
KW - artificial intelligence
KW - toxicity
KW - mutagenicity
KW - cancer
UR - http://www.scopus.com/inward/record.url?scp=85166915200&partnerID=8YFLogxK
U2 - 10.1016/B978-0-443-15274-0.50432-7
DO - 10.1016/B978-0-443-15274-0.50432-7
M3 - Article
SP - 2717
EP - 2722
JO - Computer Aided Chemical Engineering
JF - Computer Aided Chemical Engineering
ER -