Evaluation of data non-quality in database or data warehouse systems is a preliminary stage before any data usage and analysis, moreover in the context of data integration where several sources provide more or less redundant or contradictory information items and whose quality is often unknown, imprecise and very heterogeneous. Our application domain is bioinformatics where more than five hundred of semi-structured databanks propose biological information without any quality information (i.e. metadata and statistics describing the production and the management oft he biological data). In order to facilitate the multi-source data integration in the context of distributed biological databanks, we propose a technique based on the concepts of quality contract and data source negotiation for a standard wrapper-mediator architecture. A quality source contract allows to specify quality dimensions necessary to the mediator for data extraction among several distributed resources. The source selection is dynamically computed with the contract negotiation which we propose to include into the mediation and the global query processings before data acquisition. The integration of the multi-source biological data is differed for the restitution and combination of the results of the global user’s query by techniques of data recommendation taking into account source quality requirements.