Molecular evolution is an important step in the development of therapeutic antibodies. However, the current method of affinity maturation is overly costly and labor-intensive because of the repetitive mutation experiments needed to adequately explore sequence space. Here, we employed a long short term memory network (LSTM)—a widely used deep generative model—based sequence generation and prioritization procedure to efficiently discover antibody sequences with higher affinity. We applied our method to the affinity maturation of antibodies against kynurenine, which is a metabolite related to the niacin synthesis pathway. Kynurenine binding sequences were enriched through phage display panning using a kynurenine-binding oriented human synthetic Fab library. We defined binding antibodies using a sequence repertoire from the NGS data to train the LSTM model. We confirmed that likelihood of generated sequences from a trained LSTM correlated well with binding affinity. The affinity of generated sequences are over 1800-fold higher than that of the parental clone. Moreover, compared to frequency based screening using the same dataset, our machine learning approach generated sequences with greater affinity.

Base Editor, a technique that utilizes Cas9 nickase fused with deaminase to introduce single base substitutions, has significantly facilitated the creation of valuable genome variants in medical and agricultural fields. However, a phenomenon known as RNA off-target effects is recognized with Base Editor, resulting in unintended substitutions in the transcriptome. It has been reported that such substitutions often occur in specific base motifs (ACW), but whether these motif mutations are dominant has not been investigated. In this study, we constructed a pipeline for analyzing RNA off-target effects, called the Pipeline for CRISPR-induced Transcriptome-wide Unintended RNA Editing (PiCTURE), and analyzed RNA-seq data previously reported. We found minor RNA off-target effects associated with the reported base motifs, and most were indistinguishable in motif analysis.Consequently, we trained a Large Language Model (LLM) specialized for DNA base sequences on RNA off-target sequences and developed a classifier for assessing the risk of RNA off-target effects based on the sequences. When the model's estimations were applied to the RNA off-target data for BE4-rAPOBEC1 and BE4-RrA3F, satisfactory determination results were obtained. This study is the first to demonstrate the efficacy of machine learning approaches in determining RNA off-target effects caused by Base Editor and presents a predictive model for the safer use of Base Editor.