Diagnostic virus isolation is still frequently used, particularly from respiratory tract secretions. Testing positive virus cultures for all possible viruses is time-consuming, and unexpected or unknown viruses may escape detection. Therefore, a novel random PCR approach was developed that allows sequence-independent amplification of viral nucleic acids from virus isolation-positive cultures. Selectivity for viral sequences is obtained by preferential isolation of nucleic acids that are particle associated and resistant to nucleases. Using primers with a degenerated 3' end, the isolated nucleic acids are amplified and the randomly amplified PCR products are cloned and sequenced. As proof of the concept, the PAN-PCR approach was applied to supernatants of coxsackievirus B3 and murine adenovirus type 1-infected cells. Enterovirus and adenovirus sequences were obtained, demonstrating that the random PCR approach allows detection of RNA and DNA viruses. As a first application of this PAN-PCR approach, we characterized a virus isolate from mouth-washing material of a patient with chronic fatigue syndrome and high antibody titers to coxsackievirus B2. The virus isolate had tested negative for enteroviruses and respiratory viruses (influenza viruses A and B, parainfluenza virus types 1 to 3, respiratory syncytial virus, and adenovirus) by immunofluorescence and PCR. Particle-associated, nuclease-resistant RNA and DNA were prepared from the supernatant of infected cells. The DNA and the reverse-transcribed RNA were randomly amplified, and PCR products were cloned and sequenced. Of 25 sequences obtained from the DNA preparation, 24 contained herpes simplex virus type 1 (HSV-1) sequences from 14 different loci spread over the HSV-1 genome. This result was confirmed by using a standard diagnostic HSV-PCR, demonstrating that the PAN-PCR correctly identified the virus isolate. Although the identification of HSV-1 in mouth-washing material is not surprising in retrospect, it clearly demonstrates the applicability of the PAN-PCR approach. This method should be particularly useful for characterizing virus isolates that have tested negative for all expected viruses and for identifying unknown viruses.