Recently, more and more research has been focused on the analysis of polar organic compounds as they tend to be persistent and mobile in the aquatic environment. The serial coupling of reversed-phase and hydrophilic interaction liquid chromatography allows the separation of analytes of an extended polarity range within a single run. The non-target screening approach was driven by high-resolution mass spectrometry and is able to detect unexpected compounds. It is therefore capable of complementing regular monitoring of surface water. Non-target screening, however, can produce massive data sets. Here, a data processing method is presented focusing on unravelling tentative polar compounds from the full scan data of 51 samples of the Danube river and its tributaries. The feature extraction method was optimized to 34 reference compounds at two concentration levels and was then applied to real samples. Features were matched by accurate mass with anthropogenic substances stored in the compound database STOFF-IDENT located on the FOR-IDENT platform. In order to extract polar candidates, the retention time interval corresponding to the HILIC separation was connected to compounds with a negative log D value. As a result, 67 candidates were detected which were found to be plausible. Finally, features were prioritized based on an identification certainty classification system as well as their frequency of occurrence. Therefore, several feature-candidate compound pairs could be suggested for confirmation via reference materials. The presented non-target screening strategy followed by a database query is transferrable to other sample sets and other data evaluation tools.