Clustering is one of the basic tasks in machine learning and data mining. Euclidean minimum-sum-of-squares clustering problem is probably the most common clustering model. It consists in finding k cluster centers so as the sum of squared Euclidean distances from data items to their closest centers is minimized. The problem is known to be nonconvex and NP-hard even in the planner case. In this paper, we consider a generalization of the Euclidean minimum-sum-of-squares clustering problem and propose a new DC programming approach to solve it. First, following the methodology of supervised machine learning, we construct the matrix that defines the Mahalanobis distance using items side-information. Then, we cast the original nonconvex problem as a continuous optimization problem with the objective function represented as a DC function. Next, we devise a solution algorithm, resting upon the global optimality conditions and global search scheme for DC minimization problems proposed by A.S. Strekalovsky. We implemented the developed algorithm and demonstrated its competitiveness with the conventional k-means heuristic in a series of computational experiments.