The distribution of DNA molecular lesions is defined as the histogram
of lesion occurrence probability versus lesion type and is termed a mo
lecular lesion spectrum (MLS). The MLSs for 405 nm, 254 nm and gamma r
ay electromagnetic radiations are investigated here, and it is shown t
hat eight, five, four, three and two lesion sets function as very good
radiation signatures: that is, MLSs based upon these lesion sets can
be used to identify the causal electromagnetic radiation. The approach
adopted is based on pattern recognition techniques couched in neural
network formalism, the patterns being defined as vectors and their uni
queness described as a degree of vector orthogonality. The minimal les
ion set for signature of the three radiations is found to consist of t
wo lesions, thymine dimer (d) and locally multiply damaged site (lmds)
lesions. This is a surprising finding, and it works only because the
d lesion turns out to be a radiation marker for 254 nm radiation and t
he lmds lesion to be a radiation marker for gamma rays. Since these fi
ndings about signatures and markers are important from safety, insuran
ce and medical aspects, it is well to caution that only three radiatio
ns, all electromagnetic, have been investigated here. Work on particul
ate and other electromagnetic radiations is underway.