Abstract. We present the notion of sequential association rule and introduce Sequential Nuggets of Knowledge as sequential association rules with possible low support and good quality, which may be highly relevant to scientific knowledge discovery. Then we propose the algorithm SNK that mines some interesting subset of sequential nuggets of knowledge and apply it to an example of molecular biology. Unexpected nuggets that are produced may help scientists refine a rough preliminary classification. A first implementation in Java is freely available on the web3 .