Grading breast cancer malignancy with neural networks

The three-dimensional structures generated for 20 “never born proteins” (NBP – random amino acid sequence with no signifi cant homology to existing proteins) using two different techniques: ROSETTA (called R in the paper) and “fuzzy oil drop” model (called S in the paper) were compared to estimate the accordance with the assumed model estimating the infl uence of an external force fi eld on the fi nal structure of the protein. Selected structures are those corresponding to the highest (10 proteins) and lowest (10 proteins) RMS-D values obtained measuring the similarity between the R and S structures. The R structures generated according to an internal force fi eld (the individual inter-molecular interaction) including solvation effects were analyzed using the “fuzzy oil drop” model as target model. The second applied model “fuzzy oil drop” generated structures characterized by an ordered hydrophobic core structure. 13 of the 20 selected S structures appeared to be accordant with the “fuzzy oil drop” model while 6 out of the 20 structures appeared to be accordant with external force fi eld for R structures which suggests a general interpretation of the infl uence of an external force fi eld on the folding simulation.