Background A number of genetic defects in humans are due to point mutations in a single, often tightly regulated gene. Genetic treatment of such defects is preferably done by correcting only the altered base pair at the endogenous locus rather than by a gene replacement strategy involving viral vectors. Promisingly high repair rates have been achieved in some systems with the non-viral approach of transfecting chimeric RNA/DNA oligonucleotides (chimeraplasts). However, since this technique does not yet perform robustly, several parameters thought to be important in oligonucleotide-mediated gene repair were examined. Methods A series of transgenic HEK-293 cell clones has been established harboring high or low copy numbers of a point-mutated enhanced green fluorescent protein' (EGFP) gene as the target. At the level of single living repair efficiencies were measured by fluorescence-activated cell sorting cells, (FACS) regarding topology (single-stranded, double-stranded), exonuclease protection (four phosphorothioate linkages at both ends), polarity (sense, antisense), and length (13mer, 19mer, 35mer, 69mer) of the oligonucleotide. Results When targeting chromosomal loci, up to 0.2% corrected cells were obtained with single-stranded unmodified oligodeoxynucleotides, whereas a chimeraplast, its DNA analogue, and double-stranded DNA fragments were practically non-functional. Correction efficiencies correlated with target gene copy numbers. Modifying exonuclease resistance, polarity or length of single-stranded oligodeoxynucleotides did not enhance repair efficacy above the sub-percentage range. Conclusions Successful chromosomal reporter gene repair in HEK-293 cells required an oligodeoxynucleotide to be single-stranded. In concert with the gene copy number correlation, functional interaction between the repair molecule and the target site seems to be one bottleneck in targeted gene repair. Copyright (C) 2004 John Wiley Sons, Ltd.