Dear Saiful

Yes it definitely is possible and see situations all of the time where you have a predicted annealing temp of (say) 63C but a real specific product at 67C. Generally speaking, specific product is yielded any where from Tm (melting temp of primer) -5C to Tm+2C. For that reason, if you do not try and evaluate your optimal annealing temperature empirically by trying a temp gradient - normally from Tm - 2C (although sometimes going down as low as Tm-5C is necessary) to Tm+2C then most people opt for Tm-2C is a likely optimal annealing temp in terms of specificity i.e. a single band and efficiency, i.e. the amount of that specific band; In particular:

1. The determination of Tm is dependent on algorithm: Some algorithms take account of actual PCR reactions conditions like primer concentration, Sodium and Magnesium concentration and dNTP concentration all of which have an effect on duplex stability (as well as primer base composition) and therefore the actual real time annealing temp of your primer in your reaction tube. A lot of algorithms however do not ask for these conditions and just predict melting temp based on something called the 'nearest neighbour' formula which looks at the respective % of A G C and T in your primer and their distribution

In other words the algorithm you have used might (and has) predicted 65C as your melting temp (compatible with annealing at 63C) but another might predict 65C ( they tend to differ by 2-3C)

2. More importantly, primers will tend to yield specific products anywhere from Tm-5C to Tm +2C: If you perform such annealing gradient you tend to find that all can give your specific product but the efficiency drops as you go from low to high temp, particularly above the predicted Tm: In other words from low to high temp the amount of specific product you obtain decreases

Thus, in your case you might have a melting temp of (for example) 65C based on your particular algorithm which might suggest an optimal annealing temp of 63-65C. Nevertheless, PCRs invariably work at Tm +2C which based on a predicted melting temp of 65C would imply successful PCR at 67C, although you might find the amount of specific product could be less

Keep in mind that these algorithms predicting melting temp are just that - educated guesses based on thermodynamics - and the reality in your tube, based on the local environment - the purity of your target, the GC content of your target & primer, secondary structure in both your primer(s) and target,  ion composition, dNTP and Mg concentration - might actually be different. This is why there is not one hard and fast unequivocal algorithm and PCRs will always work specifically for a range of temp; Invariably Tm-5C to Tm +2C but ultimately the actual reality tends to be peculiar to each target and primer pair. These things are guides not gospel truth 

Dear Saiful

Yes it definitely is possible and see situations all of the time where you have a predicted annealing temp of (say) 63C but a real specific product at 67C. Generally speaking, specific product is yielded any where from Tm (melting temp of primer) -5C to Tm+2C. For that reason, if you do not try and evaluate your optimal annealing temperature empirically by trying a temp gradient - normally from Tm - 2C (although sometimes going down as low as Tm-5C is necessary) to Tm+2C then most people opt for Tm-2C is a likely optimal annealing temp in terms of specificity i.e. a single band and efficiency, i.e. the amount of that specific band; In particular:

1. The determination of Tm is dependent on algorithm: Some algorithms take account of actual PCR reactions conditions like primer concentration, Sodium and Magnesium concentration and dNTP concentration all of which have an effect on duplex stability (as well as primer base composition) and therefore the actual real time annealing temp of your primer in your reaction tube. A lot of algorithms however do not ask for these conditions and just predict melting temp based on something called the 'nearest neighbour' formula which looks at the respective % of A G C and T in your primer and their distribution

In other words the algorithm you have used might (and has) predicted 65C as your melting temp (compatible with annealing at 63C) but another might predict 65C ( they tend to differ by 2-3C)

2. More importantly, primers will tend to yield specific products anywhere from Tm-5C to Tm +2C: If you perform such annealing gradient you tend to find that all can give your specific product but the efficiency drops as you go from low to high temp, particularly above the predicted Tm: In other words from low to high temp the amount of specific product you obtain decreases

Thus, in your case you might have a melting temp of (for example) 65C based on your particular algorithm which might suggest an optimal annealing temp of 63-65C. Nevertheless, PCRs invariably work at Tm +2C which based on a predicted melting temp of 65C would imply successful PCR at 67C, although you might find the amount of specific product could be less

Keep in mind that these algorithms predicting melting temp are just that - educated guesses based on thermodynamics - and the reality in your tube, based on the local environment - the purity of your target, the GC content of your target & primer, secondary structure in both your primer(s) and target,  ion composition, dNTP and Mg concentration - might actually be different. This is why there is not one hard and fast unequivocal algorithm and PCRs will always work specifically for a range of temp; Invariably Tm-5C to Tm +2C but ultimately the actual reality tends to be peculiar to each target and primer pair. These things are guides not gospel truth 

If you get clearer results you are on the right path. The optimal annealing temperature depends primarily on length and base composition of the PCR primer and can vary from 50 C to 72 C, or even beyond. Also salt concentrations affect the Melting temperature (Tm). Increasing the annealing temperature decreases the chance of unspecific PCR products, but if too high nothing is amplified.

There are numerous web calculators for optimization of annealing temperatures and for designing primers pairs with similar Tm's, below are links to a few of them. Don't be suprised if they give dissimilar temperatures, that is because they calculate Tm's using different algorithms.

http://tmcalculator.neb.com/#!/

http://biotools.nubic.northwestern.edu/OligoCalc.html

http://tinyurl.com/z546ofb

Well said Peter

I concur and in fact use Oligo calculator myself

I would add that in addition - whilst impinging on the general subject of primer design -  I also use the oligo design tool by IDT to screen for hetero or primer dimers:

http://www.idtdna.com/calc/analyser

In essence, I think oligo calculator is proficient when it comes to intra primer annealing including hair pin loops

IDT primer design tool identifies hetero (primer) dimers more effectively but crucially in the context of the original question allows you to simulate reaction conditions and thus the likely real Tm in the context of your local PCR reaction

Thank you Mr. Laurence & Peter

I agree with you. After I use 67C, the specific product becomes clearer, but decreased. How to increase the specific product after that? Add the DNA product? But sometimes appears one or more unspecific band/products.

Dear Saiful

No adding more product is probably a bad idea. Assuming the number of PCR cycles you have is between 25 and 35, the most effective way to increase specific product is to add another 5 cycles to your PCR; That is if you currently use 30 cycles go upto 35; if you currently use 35 cycles go up to 40. Generally speaking it is not a good idea to go higher than 40 cycles as you increase the risk of amplifying other non specific bands

There are other things you can do but I would start by adding x5 more cycles to your PCR; 1, 2 cycles extra incidentally will not make a substantial difference to the amount of PCR product you amplify

Ok Mr Laurence. I will try it. Thank  you

I agree with Laurence, adding DNA most likely makes things worse as there will be less time for the polymerase to extend the primers and in this situation the longer product suffers first. However, the situation would be similar with added cycles as there at some point will be more product acting as template. By adding cycles both products will increase, but probably the shorter amplicon benefits more. But if your main concern is to get a readable signal from both this might not be an issue.

If adding extension time is not helpful enough, a way to give more competetive edge to the longer amplicon is to redisign the primers so that tghe longer amplicon primers have a higher Tm. In that case primer annealing occurs earlier as the temperature drops from denaturation to annealing temperature and consequently primer extension starts earlier.

I generally concur again with Peter

extending Peters point in general for standard Taq you require 1 minute per kb of extension time and 2 min per kb with proof reading polymerases. This will be most effective in producing more product at the right annealing temp

However, I have found in my experience that increasing cycle number of increments of 5 cycles up to 40 cycles can be helpful

Laurence is correct about extension time of Taq (1 min per kb). With proof reading polymerases there is variation. When using polymerases that are fused with a DNA binding protein processitivity increases dramatically. Such polymerases include Phusion and Q5, for which 15-30 sec per kb is enough.

In Saiful's case adding cycles combined with less template DNA would probably improve results. Using above mentioned Phusion or Q5 would probably also help and  total PCR time should not increase even with added cycles as extension time can be shortened. I haven't tried Q5 yet, but have found Phusion very good for duplex and multiplex PCR. In their respective buffers, both systems work best by increasing annealing temperature by about 3 C and in Saiful's case this would make a two-step PCR (with 70 C as annealing/extension) possible.

Peter is right about the newer polymerases. I have used phusion extensively and most of the time string specific product is yielded by much lower extension times as stated

I was making a generalisation in terms of requiring extra time for proof reading principles but yes you are right Peter those absolute times are more applicable 10 years ago, when dealing with the old pfu and HiFi enzymes and newer generation polymerases, exemplified by phusion, are highly accurate fast and processive so times stated are correct for these enzymes. I have used both sets of proof readers so I speak from personal experience when I say this speeding up is true