Predicting the Expression Levels of Recombinant Protein by using Reinforcement Learning

S. Kira, M. Yamamura, and S. Kuramitsu (Japan)


Recombinant protein expression, protein translation, rein forcement learning, data mining, bioinformatics.


Recombinant protein expression is an important technique to product useful proteins. The expression system uses host cells, which are bacteria, yeast and mammalian cells etc. Target genes coded useful proteins are transformed into the cells and then the cells express the target protein. How ever, the expression ef´Čüciencies of the target protein dif fer vastly depending on the sequence properties, unfortu nately the targets often do not express. So we proposed the prediction method of the expression levels of recombinant protein to improve the expressions based on Reinforcement Learning (RL). Our approach used gene sequences whose expression levels were known from the past investigations. We focused the protein translation process and described the process as markov decision process (MDP). MDP pa rameters were adjusted to identify the expression levels of the genes by using RL algorithm. We applied our algorithm to real world data. Using the learning result, we simulated the protein translation for a new target gene and compared the real expression levels of the gene. The simulation result agreed with the actual expression level of the gene.

Important Links:

Go Back