Host: The Japanese Society of Toxicology
Name : The 51st Annual Meeting of the Japanese Society of Toxicology
Date : July 03, 2024 - July 05, 2024
[Introduction]Gapmer antisense oligonucleotides (Gapmer) are chemically modified with artificial nucleic acid at both ends. Some Gapmers have hepatotoxicity, and it has been reported a relation between base sequence and chemical modification. Therefore, in this study, we focused on the increase in blood alanine aminotransferase (ALT), an indicator of hepatotoxicity, in animal tests, and aimed to develop a machine learning model to predict ALT elevations from base sequences and chemical modification information.[Method]Data on Gapmers’ sequences, modifications, ALT, and doses were collected from papers on single-dose studies in mice. The objective variable was the presence or absence of ALT elevation. A positive substance was defined as a substance that showed an increase of 10-fold or more of ALT value compared to the control group, and a negative was defined as an increase of less than 10-fold. As explanatory variables, each sequence was converted to the frequency of 2 or 3-base subsequences. The sugar modification structure information was calculated from alvaDesc. A random forest classification model was constructed.[Result/Discussion]We obtained data on 138 Gapmers from 16 studies. Of these, 86 were positive substances and 52 were negative. The model performance was ROC-AUC: 0.84, sensitivity: 0.88, and specificity: 0.55. In this study, we collected data on ALT in single-dose studies of Gapmers and developed a high-performance prediction model using only the substance information. This model is expected to support the efficient design of Gapmers to avoid ALT elevation.
