SPINE: an integrated tracking database and data mining approach for identifying feasible targets in high-throughput structural proteomics

High-throughput structural proteomics is expected to generate considerable amounts of data on the progress of structure determination for many protein s. For each protein this includes information about cloning, expression, pu rification, biophysical characterization and structure determination via NM R spectroscopy or X-ray crystallography. It will be essential to develop sp ecifications and ontologies for standardizing this information to make it a menable to retrospective analysis. To this end we created the SPINE databas e and analysis system for the Northeast Structural Genomics Consortium. SPI NE, which is available at bioinfo.mbb.yale.edu/nesg or nesg.org, is specifi cally designed to enable distributed scientific collaboration via the Inter net. It was designed not just as an information repository but as an active vehicle to standardize proteomics data in a form that would enable systema tic data mining. The system features an intuitive user interface for intera ctive retrieval and modification of expression construct data, query forms designed to track global project progress and external links to many other resources. Currently the database contains experimental data on 985 constru cts, of which 740 are drawn from Methanobacterium thermoautotrophicum, 123 from Saccharomyces cerevisiae, 93 from Caenorhabditis elegans and the remai nder from other organisms. We developed a comprehensive set of data mining features for each protein, including several related to experimental progre ss (e.g. expression level, solubility and crystallization) and 42 based on the underlying protein sequence (e.g. amino acid composition, secondary str ucture and occurrence of low complexity regions). We demonstrate in detail the application of a particular machine learning approach, decision trees, to the tasks of predicting a protein's solubility and propensity to crystal lize based on sequence features. We are able to extract a number of key rul es from our trees, in particular that soluble proteins tend to have signifi cantly more acidic residues and fewer hydrophobic stretches than insoluble ones. One of the characteristics of proteomics data sets, currently and in the foreseeable future, is their intermediate size (similar to 500-5000 dat a points). This creates a number of issues in relation to error estimation. Initially we estimate the overall error in our trees based on standard cro ss-validation. However, this leaves out a significant fraction of the data in model construction and does not give error estimates on individual rules . Therefore, we present alternative methods to estimate the error in partic ular rules.