Fig 1.
A schematic diagram of in vivo protein half-life prediction from cellular properties.
Fig 2.
Linear regression between the half-lives of proteins present in the murine liver tissue and cell culture (e.g. NIH3T3[2]) data sets.
Fig 3.
Common liver protein subset with the highest correlation coefficient between the protein half-life in the tissue and cells.
Fig 4.
The clustering scheme of common liver protein data sets.
The protein clusters form from the linear regression line (Fig 2). C1 has very long-living half-life proteins while others (C2, C3) have short-living proteins.
Table 1.
Coefficients of the linear regression between the common liver protein half-life in the tissue and cell of each cluster (Fig 4).
Table 2.
Performance analysis of the protein half-life prediction of the model with two types of protein properties (e.g. PCH, ACH) using two-thirds (e.g. ()) of total data sets (common liver proteins) of each cluster.
All protein properties are designated by ACH, and positively-correlated proteins are designated by PCH. The best result provided is the C2 cluster. It has predicted 44% of protein half-lives between 10% deviation from the experimental value.
Fig 5.
Artificial neural network between the protein properties and clusters.
Fig 6.
Uncommon protein half-life prediction with noise.