I've been told that the maximum number of cycles in PCR is between 20 and 30.
Is this true, and what are the reasons for this limitation?
I would draw the line beyond 35, but thats a bit cosmetic. The reasons are manyfold:
If you need a higher sensitivity with more cycles, you can use the technique called "nested PCR". There you do a first round with a primer pair specific for the region of interest and then do a second round with primers which are located slightly to the inside of the amplified DNA. This is done to avoid the amplification of unwanted contaminations. Since you do some 50-70 rounds of PCR amplification in total, this method is extremely sensitive (also to contaminations). See the image from the Wikipedia article for details:
How many cycles of PCR before dNTPs run out?
Assume a 25 μl reaction.
Assume 200 μM dNTPs.
200 μM dNTPs = 200 pmol μl -1
so in 25 μl reaction, there are 5000 pmol of dNTPs
= 5000 x 10-12 x 6 x 1023 molecules
= 3 x 1015 molecules dNTP
Assume that we start with 1 molecule of a 1000 bp template, 50% GC
1 kb = 2000 nucleotides
So , how many of these molecules can we construct using 3 x 1015 molecules dNTP?
= 3 x 1015/2000
= 1.5 x 1012 DNA molecules
How many cycles of PCR to produce this many from a single template molecule?
2n = 1.5x 1012
nlog2 = log(1.5) + log(1012)
nlog2 = 12.18
n = 12.18/ log2 = 41 cycles
Of course this in an absolute upper limit. The estimate assumes that you start with one template molecule of 1 kb, and that dNTPs aren't being hydrolysed, or otherwise degraded.
In my own research I've found that the amount of PCR product increased as I increased the cycles to 50 times. This was for a single pair of primers used to amplify genomic DNA extracted with a particular kit as doesn't apply to other reactions I've done. What some people seem to forget is that the PCR products don't necessarily double at each cycle at any point in the PCR because some primers don't function as well as others and reaction conditions may be suboptimal (e.g. presence of contaminants or concentration of Mg ions).
In short, provided you aren't getting reduced-quality products (e.g. smearing or non-specific priming products), then I think it's fine to increase the number of cycles well beyond 30 - it's something you have to test empirically with your particular primers, template, polymerase, buffer etc.