Since SABIO-RK stores information about reactions and their kinetic properties and in addition experimental conditions under which kinetic parameters were measured we also had a closer look at the correctness and completeness of the assay conditions because temperature, pH-value and the buffer composition are essential for the interpretation of experimental results. About 10% of the analyzed papers contain

no information about the temperature used in the experiments. About 3% of the papers only give the imprecise information that the experiments were done at “room temperature”. In about 10% of publications learn more the authors refer to another paper for the experimental method used for the measurement which causes a time-consuming search for the correct method in a reference paper. Sometimes the reference paper again refers to another paper for the method description. 20% of the publications describe the buffer composition and the compound concentrations not in standard units but use an indication of weight per assay volume which has to be manually converted to a standard unit. A biochemical reaction is defined by the chemical compounds as reaction participants in particular substrates, products, enzymes and reaction modifiers like inhibitors and activators. About 25% of the publications used for insertion in SABIO-RK only

contain incomplete reaction descriptions. For example in many cases the corresponding product pentoxifylline for a substrate used in the experimental assay is missing. For RG7420 chemical structure data insertion in SABIO-RK

biochemical reactions have to be complete containing all substrates and products. If the corresponding product information is missing in the publication SABIO-RK curators have to deduce the product(s) manually or if not possible include Unknown as compound. Frequently kinetic parameters in a paper were compared with values from other publications and were represented together in one table. Then the legend of the table or some phrases in the free text refers to the original source. Our analysis shows that there are no standard guidelines for authors how to refer to referenced values. The challenge for data extraction is here to filter the reference values from the original paper values. In SABIO-RK the parameter values are always only linked to the original source. The examples for the challenges of correct data extraction from the literature as mentioned above illustrate that a large amount of manual work by experts in biology is still needed. Natural language processing tools for automatic data extraction and text understanding are far away from being suitable for our application. Ideally journal editors should ask the authors for complete, standardized and structured data in their future articles. Collaborations between the publisher and the database site to develop common standards and data format are preferable.