Sentiment Analysis for RRI Disease Prediction - A Comparative Review

The rapid growth of biomedical literature over the past decade has created an urgent need for intelligent computational frameworks capable of extracting structured knowledge and implicit sentiment from scientific texts. In the context of RNA-RNA interactions (RRIs), a substantial portion of biologically and clinically relevant information-such as interaction confidence, experimental support, and disease relevance-remains embedded in unstructured literature rather than curated databases. Recent advances in Natural Language Processing (NLP), particularly Transformer-based models such as BERT, have enabled deeper contextual understanding of biomedical text and opened new avenues for sentiment-aware literature mining. Domain-adapted models including RIScoper, BioBERT, SciBERT, and PubMedBERT offer different levels of domain specialization and capabilities for biomedical text analysis. This review provides a comparative analysis of these models for RRI-related literature mining and sentiment analysis, examining their architectures, training corpora, domain specialization, and practical applications. The review further discusses their strengths, limitations, and potential contributions to interaction prioritization, disease association analysis, and hypothesis generation. By synthesizing advances in biomedical NLP and RNA interaction biology, this work highlights the potential of sentiment-aware text mining to augment computational RRI prediction and accelerate discovery in RNA-centric disease research.