Avian influenza virus (AIV) can directly cross species barriers and infect humans with high fatality. Using machine learning methods, the present paper scores the amino acid mutations and predicts interspecies transmission. Initially, 183 signature positions in 11 viral proteins were screened by the scores of five amino acid factors and their random forest rankings. The most important amino acid factor (Factor 3) and the minimal range of signature positions (50 amino acid residues) were explored by a supporting vector machine (the highest-performing classifier among four tested classifiers). Based on these results, the avian-to-human transmission of AIVs was analyzed and a prediction model was constructed for virology applications. The distributions of human-origin AIVs suggested that three molecular patterns of interspecies transmission emerge in nature. The novel findings of this paper provide important clues for future epidemic surveillance.
By successively assembling genetic parts according to grammatical models, complex genetic constructs can be built. However, every category of genetic parts includes many parts. With the increasing quantity of genetic parts, the process of assembling a few sets of genetic parts can be costly, time consuming, and error prone. At the final assembly step, it is difficult to decide which part should be selected. Based on a statistical language model, a dynamic programming algorithm was designed to solve this problem. The algorithm optimizes the results of genetic designs and finds an optimal solution. In this way, redundant operations can be reduced, and the cost for assembling can be minimized.