A mathematical model for secondary structure in proteins

We propose a new mathematical model for secondary structure in proteins. Our model is inspired by percolation theory on binary strings. What sets us apart from similar work on the subject is our attempt to deviate from a data mining approach (which is mostly the trend is science these days). Therefore, in predicting secondary structures, we make it our challenge to adhere to sequence information alone, in a non ad-hoc way, with only minimal information extracted from databases of known structures. Initial results show that our model captures some essential aspects of structure formation, notably a de novo discovery of hydrophobicity from an optimization perspective. A comparison of our prediction algorithm to similar methods shows improved performance. In addition, some evolutionary algorithms using our model exhibit convergences that are consistent with information obtained from structural biology.