DNA replication is more accurate than transcription (or RNA replication) because mechanisms exist for proof-reading and repair of DNA, but not for RNA. Consider a segment of a DNA strand, AGTC. Its complement is GACT. Now suppose its complement is mutated to TACT — the DNA repair system will replace the wrong T by G. Why isn’t A in the original strand replaced by C? How does the system determine that the first strand is correct and its complement is incorrect?
The reason the fact it isn't realistic is important. DNA repair machinery works by repairing common errors that occur due to common mutational pathways. The enzymes are specific for this, for example one particular enzyme targets mutations caused by UV and itself is activated by sunlight thus it's specificity makes it repair the correct chain. Also during replication, newly synthesised DNA lacks methylation. The repair enzymes thus know which chain is correct.
There are many types of DNA repair. Let us limit to the following three different types to answer this condition.
1. UV damage
Exposure of a cell to UV light can result in the covalent joining of two adjacent pyrimidines (usually thymines), producing a dimer. These thymine dimers prevent DNA polymerase from replicating the DNA strand beyond the site of dimer formation.
In such cases, UV-damage specific endonucleases cleave both the 5' and the 3' end of the lesion. Along, with the lesion some normal DNA is also lost as an oligonucleotide. Here modifications are being made only on one strand, the other is taken as a template (correct) and the DNA in the lost part is replaced.
The UV-damage may occur on both the strands but locally, one is affected and the other is the template (since you can't have thymine both sides)
2. Base Excision
Normal DNA bases can get deaminated spontaneously or on insult to an abnormal base (not on DNA).The abnormal base is excised by glycosylases and gap is later filled by considering the other strand the template.
3. Mismatch repair
What you are talking of in your example is not a case of deamination or UV damage but a mismatch repair. There can be errors in replication too where such a case could set in. There are specific enzymes for this repair, but discrimination is based on the degree of methylation. GATC sequences, which are found approximately once every thousand nucleotides, are methylated on the adenine residue. This methylation is not done immediately after synthesis, so the newly synthesized DNA is hemimethylated (that is, the parental strand is methylated but the daughter strand is not). The methylated parental strand is assumed to be correct, and it is the daughter strand that gets repaired.
The above said mechanism of discrimination is only in the gram negative bacteria. To quote Wikipedia,
However, in other prokaryotes and eukaryotes, the exact mechanism is not clear. It is suspected that, in eukaryotes, newly synthesized lagging-strand DNA transiently contains nicks (before being sealed by DNA ligase) and provides a signal that directs mismatch proofreading systems to the appropriate strand. This implies that these nicks must be present in the leading strand, and evidence for this has recently been found.
There are other cases where there is damage occuring to both strands that I'd suggest for further reading.
- Lehninger's Principles of Biochemistry, Ed. 6
- Lippincott's Illustrated Reviews, Biochemistry, Ed. 5
- Wikipedia. DNA mismatchrepair
Your specific scenario is not realistic, thymine is a pyrimidine base while guanine is a purine base. They are very different chemically and can't spontaneously convert into each other. So this is a type of damage that only occurs during the replication of DNA.
The typical damage that occurs and can be repaired is oxidation, methylation and hydrolysis of DNA. This kind of damage results in modified or missing bases and can be detected due to that.
You might also be interested in why thymine is used in DNA instead of uracil, as uracil can be easily damaged in a way that would not be detectable.
To solve the problem of which strand is edited to correct mistakes, immediately after DNA replication, the original strand is methylated to prevent correction on this strand.Thus only the newly synthesized strand is searched for errors for correction. In short methylation prevents the template strand from being edited error and correction.