Given the increasing number of biological sequences stored in databases, there is a large source of information that can benefit several sectors such as agriculture and health. Machine Learning (ML) algorithms can extract useful and new information from these data, increasing social and economic benefits, in addition to productivity. However, the categorical and unstructured nature of biological sequences makes this process difficult, requiring ML expertise. In this paper, we propose and experimentally evaluate an end-to-end automated ML-based framework, named BioPrediction, able to identify implicit interactions between sequences, e.g., long non-coding RNA and protein pairs, without the need for end-to-end ML expertise. Our experimental results show that the proposed framework can induce ML models with high predictive accuracy, between 77% and 91%, which are competitive with state-of-the-art tools.