 |
 |
 |
|
|
Need for Quantitative, Predictive Models
|
|
|
| The hybridization properties of a primer or probe are critical to its success or failure. Software must be able to predict these properties in order to select the optimal primer or probe from among tens, hundreds, thousands, or tens of thousands of candidates. Prediction is necessary because nearly every candidate has a unique sequence, and therefore a unique secondary structure, melting temperature, and potential for off-target hybridization. Direct, experimental determination of hybridization properties, even for a small fraction of candidates, would be highly cost and time prohibitive. |
|
|
|
|
 |
 |
|
|
|
|
|
|
|
|
Nearest-Neighbor Modeling
|
|
|
| The nearest-neighbor model, and extensions of it for modified chemistries, is an accurate and precise method for prediction of hybridization strength and specificity for any sequence.
Determining model parameters requires the design of several hundred duplexes, which requires expertise and experience. After the design and manufacture of each duplex, its UV absorbance is measured as a function of temperature. This provides a melting temperature, but more importantly the thermodynamics (entropy, enthalpy, and free energy) of hybridization can be obtained by automated fitting to all of the data.
Thermodynamic data collected for many individual sequences is then aggregated using Singular Value Decomposition to provide the small set of nearest-neighbor parameters that gives the most compact description of all of the sequences. Such parameter sets generated for DNA are widely available, but they do not describe modified chemistries. However, given an amount of training data for any modified chemistry that is both affordable and rapid to generate, the nearest neighbor model is readily extended to provide predictions for any modified oligonucleotide.
For example, data we obtained for hundreds of oligonucleotides containing Locked Nucleic Acid (LNA) provided a set of just 16 enthalpies and entropies for the dinucleotides of the form A-LNA A, A-LNA C, etc. (McTigue et al., Biochemistry 2004). This parameter set enables quantitative predictions for the hybridization properties for any duplex with a single LNA. The work is currently being extended to mismatched sequences, multiple incorporations, and so forth. |
|
|
|
|
|
|
|
|
|
|
 |
|
 |
| Celadon, in partnership with Dr. Jason Kahn of the Department of Chemistry and Biochemistry at the University of Maryland College Park, is expert in all aspects of modified nucleic acid chemistries. Together, they have the laboratory facilities and informatic infrastructure to rapidly generate predictive algorithms for any modified chemistry. |
|
|
|