Publication

Abstract : Messenger Ribonucleic acid (mRNA) vaccine faces a challenge of structural instability, due to which the production of vaccine becomes a big challenge. The sequence information of the mRNA vaccine can provide possible degradation sites. Recently, Deep learning areas like Natural Language Processing have shown great promise in understanding these sequences. An appropriate sequence to vector representation is necessary to apply deep learning methods effectively. In this paper, pre-trained dna2vec, rna2vec, and lshvec gene embeddings are compared to identify the best vector representation for predicting the amount of degradation given the mRNA vaccine sequences. The comparison shows that dna2vec embedding performs best.