DNA melting temperature
This article gives some background on how the melting temperature of DNA is calculated. If you need assistance with the design of oligonucleotides, or if you need automated bioinformatics analyses of DNA sequence data, including prediction or analysis of its physical properties, please contact us. We will be happy to help.
For molecular biology applications such as PCR, sequencing or microarrays, it is important to determine the melting temperature of DNA, or Tm. The Tm is defined as the temperature in degress Celsius, at which 50% of all molecules of a given DNA sequence are hybridized into a double strand, and 50% are present as single strands.
Note that ‘melting’ in this sense is not a change of aggregate state, but simply the dissociation of the two molecules of the DNA double helix.
If a homogenous solution of identical double-stranded DNA molecules is heated, the strands dissociate increasingly at higher temperatures:
In the example above, the Tm would be 60°C.
Note that the process of DNA hybridization and dissociation is complex and highly dynamic. At the Tm, double strands are constantly formed and broken up, and each DNA molecules takes part in a large number of interactions over time.
Factors that affect the Tm
The melting temperature correlates with how easily a double stranded DNA complex is formed. This, in turn, depends on the likelihood of each pair of nucleotides to bond. The hybridization process is collaborative, in the sense that already formed nucleotide pairs help their neighbouring nucleotides to pair as well. This is the reason why the melting diagramm is sigmoid and not linear: The interaction between neighbouring nucleotides makes the melting and hybridization an all-or-nothing process (to the first approximation; as you can see from the diagram, the slope of the curve is not very steep).
The Tm is affected by a number of factors:
- Concentration of DNA
- Concentration of ions in the solution, most notably Mg+ and K+
- DNA sequence
- Length of DNA
The concentration of ions affects Tm because DNA is electrically charged. Therefore it interacts with ions, which can compensate this charge.
The most complex factor is the sequence of the DNA. The sequence has an impact on the Tm for a number of reasons:
- The nucleotide pair ‘A-T’ has a weaker bond than the nucleotide pair ‘G-C’
- Nucleotides on the same strand can interact with each other, forming so-called secondary structures such as internal loops; these structures compete with the formation of the double helix and can thus increase the Tm
- Neighbouring nucleotide pairs can interact with each other. It is energetically favourable for nucleotide pairs to be neighboured with other nucleotide pairs. This so-called stacking effect decreases the Tm
How to predict the Tm
There are several methods to calculate a theoretical Tm, based on different physical models of what is happening in the hybridization or melting process. Ranked from simplest to most accurate, the methods are:
2+4 rule of thumb
This very simple method assigns 2°C to each A-T pair and 4°C to each G-C pair. The Tm then is the sum of these values for all individual pairs in a DNA double strand. This takes into account that the G-C bond is stronger than the A-T bond. Note that the 2+4 rule is valid for a small length range only, about 20-40 nt.
It is very easy to compute, but is of course very inaccurate. Where possible, it should be avoided.
A more sophisticated method is the linear regression based on the length of the DNA molecule and the GC ratio. the GC ratio is the number of G or C nucleotides divided by the total length of the DNA. Based on empirical data, a number of linear regression terms for the Tm have been proposed.
One term, from Bolton and McCarthy, PNAS 84:1390 (1962), as presented in Sambrook, Fritsch and Maniatis, Molecular Cloning, p 11.46 (1989, CSHL Press), is:
Tm = 81.5 + 16.6(log10([Na+])) + .41*(%GC) – 600/length
where [Na+] is the molar sodium concentration, (%GC) is the GC ratio, and length is the length of the sequence.
A similar formula is used by the prime primer selection program in GCG (http://www.gcg.com), which instead uses 675.0 / length in the last term (after F. Baldino, Jr, M.-F. Chesselet, and M.E. Lewis, Methods in Enzymology 168:766 (1989) eqn (1) on page 766 without the mismatch and formamide terms). The formulas here and in Baldino et al. assume Na+ rather than K+. According to J.G. Wetmur, Critical Reviews in BioChem. and Mol. Bio. 26:227 (1991) 50 mM K+ should be equivalent in these formulae to 200 mM Na+.
Note that these formulae are just approximations, as they do not take into account stacking effects and consider nucleotide properties only in the form of an averaged GC ratio.
Nearest Neighbour Method
A more sophisticated method – and the current state of the art – is the nearest neighbour method, which takes into account both the enthalpy of the pair formation between two nucleotides as well as the stacking effect between nearest neighboured nucleotide pairs.
The method has been published by Rychlik, Spencer, Rhoads in Nucleic Acids Research, vol 18, no 21, page 6410.
Unfortunately, as opposed to the previous two methods, this method cannot be easily computed manually or with a pocket calculator:The nearest neighbour method goes through each nucleotide position, looks up the corresponding enthalpy and entropy values for the given pair and for its interaction with the previous pair in a precalculated table, and adds them up.
The thermodynamic tables have to be assembled from meticulous calorimetric experiments, using carefully designed oligonucleotide probes. From time to time, new and improved tables are published. The currently best publicly available tables are those from SantaLucia, J. (1998) Proc. Nat. Acad. Sci. USA 95, 1460-1465.
Theoretically and ideally, the Tm would be determined by a complete simulation of the hybridization process. This would require the consideration of secondary structure formation within a single strand, the interaction between strands other than perfect hybridization (such as partial hybridization with a different part of the opposing strand) and simulation of detailed ion effects. However, this is computationally too expensive and impractical.
Therefore, it is important to bear in mind that any current method for the theoretical determination of Tm is an approximation, and results can differ significantly from the actual hybridization behaviour observed in the lab. Also, minute effects of the exact experimental setup, which are hard to control – such as the exact concentration of various ions – would make it virtually impossible to reproduce a perfect theoretical prediction in an actual experiment.
It is very important to be careful when comparing Tm’s from different calculations, such as the Tm given by an oligonucleotide synthesis company and that given by an online web-based primer design tool. Besides the various approaches, there are a number of parameters in each approach, especially DNA and ion concentration, that affect the calculated Tm. Therefore, Tm’s for the same DNA sequence can vary widely.
If you don’t know the exact method and parameters used in a Tm calculation, you should not compare the results to other calculations.
Tm calculations, therefore, are primarily useful to compare a given set of DNA sequences with each other, using the same method and parameters for all members of the set. In this way, the TM prediction can be used to determine whether a set of primers can be used with the same PCR conditions, for example, or whether microarray probes are compatible with each other.
In addition, the theoretical Tm is a very rough estimate of the optimal temperature for hybridization and a good starting point for experiments. If in doubt, it is adivsable to calculate the Tm according to more than one method and for a range of ion and DNA concentrations, to see how critical the method and parameters are.
We are enthusiastic about bioinformatics, and if you need help in designing or analyzing PCR templates, oligonucleotide probes, or primers, we suggest you check out our bioinformatics section or send us your questions. If you need a customized synthetic gene, you have come to the right place – we were one of the first providers of gene synthesis in 1999, and our advanced gene optimization and gene synthesis service is still our core business.