When shotgun sequencing projects are almost complete, continued random sequencing becomes increasingly less fruitful. The sequence may have regions of data that have only been determined on one strand (and therefore potentially error prone) or there may be gaps in the sequence. In both of these circumstances, it can be of benefit to be able to determine the sequence from the opposite strand of an Ml3 clone, or to be able to get sequence information from the opposite end of a large insert (1 ,2 ).