The ability to identify an unknown protein based solely on its composition was again the theme of this year's study. Unlike last year's study, where participating laboratories were asked to identify a soluble protein; this year, study participants were asked to identify a protein electro-blotted to PVDF membrane. Last year 90% of core labs were able to identify the protein, this year the number dropped to 73% for the best group of samples (the 10 mg level). The difference is due to the higher average percent error encountered by most laboratories with samples blotted to PVDF membranes, about 16% versus 12% for the soluble sample.
The Committee prepared a sample containing 5 mg of a known protein, ovalbumin, and three levels of an unknown protein, 1 mg, 5 mg, and 10 mg. The unknown protein was either lysozyme or BSA. The samples were electrophoresed in Laemmli gels and electroblotted to PVDF membrane using MES buffer. The membranes were stained with Coomassie blue and sent to participating core laboratories with suggestions for hydrolysis conditions.
We received 38 sets of data, about half from laboratories using postcolumn derivatization and the rest from laboratories using precolumn derivatization methods. Preliminary analysis of the data shows that the average percent error values were similar for both groups, 14.7 + 5.3% for the postcolumn methods versus 18.6 + 8.6% for the precolumn methods for the ovalbumin sample at 5 mg. The overall percent error of analysis decreased as the quantity of the sample applied to the gel increased. Overall errors for the 1 mg, 5 mg, and 10 mg samples were 31%, 20%, and 18%. This phenomenon is probably due to high background levels of amino acids seen on blank portions of the PVDF membranes. Particularly problematic amino acids were glycine, methionine, and proline.
The Committee submitted the data from these analyses to the ExPASy and the Propsearch Web sites to ensure that the same search criteria were used for all data sets. The correct protein (not necessarily from the correct species) was identified by approximately 73% of the facilities at the 10 mg level, by 53% at the 5 mg level, and by only 28% at the 1 mg level. We saw a strong correlation between the accuracy of the analysis and the ability to correctly identify the protein. In general, analyses with an average percent error of 15% or less produced data that permitted correct identification of the protein and with high confidence scores. Analyses with errors over 20% usually did not allow correct identification of the protein. Analyses with errors falling between these groups were partially successful but suffered from low confidence scores. These results suggest a target average compositional accuracy of 15% or better in order to reliably identify proteins transferred to PVDF.
We are finalizing our study for presentation at the Protein Society meeting and will provide a copy of our poster to the membership on the ABRF WWW Homepage.
Return to the The ABRF Home Page