A new model of DNA computation that is based on surface chemistry is studied. Such computations involve the manipulation of DNA strands that are immobilized on a surface, rather than in solution as in the work of Adleman. Surface-based chemistry has been a critical technology in many recent advances in biochemistry and offers several advantages over solution-based chemistry, including simplified handling of samples and elimination of loss of strands, which reduce error in the computation. The main contribution of this paper is in showing that in principle, surface-based DNA chemistry can efficiently support general circuit computation on many inputs in parallel. To do this, an abstract model of computation that allows parallel manipulation of binary inputs is described. It is then shown that this model can be implemented by encoding inputs as DNA strands and repeatedly modifying the strands in parallel on a surface, using the chemical processes of hybridization, exonuclease degradation, polymerase extension, and ligation. Thirdly, it is shown that the model supports efficient circuit simulation in the following sense: exactly those inputs that satisfy a circuit can be isolated and the number of parallel operations needed to do this is proportional to the size of *Corn, Fei, Frutos, Guo, Liu, Smith and Thiel are in the Chemistry Department, Condon and Glazer are in the Computer Sciences Department and Cai and Lagally are in the Materials Sciences Department. Email address for further communication: condon@cs.wisc.edu. Supported by NSF grant numbers CCR-9628814 and CCR-9613799. Permission to make digital/hard copies of all or part of this mnterinl for personal or classroom use is granted without fee provided that the copies are not made or distributed for profit or commercial advantage, the copyright notice, the title ofthe publication and its date appear, and notice is given that copyright is by permission of the ACM, Inc. To copy othenvise, to republish, to pod on servers or to redistribute to lists. requires specific permission .Ifld/or fee. RECOMB 97, S.anla Fe New Mexico USA Copyright 1997 ACM O-89791-882-8/97/01 ..$3.50 the circuit. Finally, results are presented on the power of the model when another resource of DNA computation is limited, namely strand length.