Base call accuracy estimates

THE USE OF BASE CALL ACCURACY ESTIMATES OR CONFIDENCE VALUES... 

Bonfield, J. K., and Staden, R. (1995).The application of numerical estimates of base calling accuracy to DNA sequencing projects. Nucleic Acids Res. 23, 1406—1410. 

Order of magnitude estimate also called ratio estimate . This method considers economic data for reference equipment and supposes that required pieces of equipment are very similar to the reference ones (in size range, materials, temperature and pressure operating range, etc.). The order of magnitude estimate uses ratios based upon specific size or equipment capacity. For such an estimate, the accuracy is about 40-50%. 

The accuracy of an analytical method is given by the extent by which the value obtained deviates from the true value. One estimation of the accuracy of a method entails analyzing a sample with known concentration and then comparing the results between the measured and the true value. The second approach is to compare test results obtained from the new method to the results obtained from an existing method known to be accurate. Other approaches are based on determinations of the per cent recovery of known analyte spiked into blank matrices or products (i.e., the standard addition method). For samples spiked into blank matrices, it is recommended to prepare the sample at five different concentration levels, ranging over 80-120%, or 75-125%, of the target concentration. These preparations used for accuracy studies usually called synthetic mixtures or laboratory-made preparations . 

Statistical estimation uses sample data to obtain the best possible estimate of population parameters. The p value of the Binomial distribution, the p value in Poison s distribution, or the p and a values in the normal distribution are called parameters. Accordingly, to stress it once again, the part of mathematical statistics dealing with parameter distribution estimate of the probabilities of population, based on sample statistics, is called estimation theory. In addition, estimation furnishes a quantitative measure of the probable error involved in the estimate. As a result, the engineer not only has made the best use of this data, but he has a numerical estimate of the accuracy of these results. 

Physical Models versus Empirical Models In developing a dynamic process model, there are two distinct approaches that can be taken. The first involves models based on first principles, called physical or first principles models, and the second involves empirical models. The conservation laws of mass, energy, and momentum form the basis for developing physical models. The resulting models typically involve sets of differential and algebraic equations that must be solved simultaneously. Empirical models, by contrast, involve postulating the form of a dynamic model, usually as a transfer function, which is discussed below. This transfer function contains a number of parameters that need to be estimated from data. For the development of both physical and empirical models, the most expensive step normally involves verification of their accuracy in predicting plant behavior. 

Autocorrelation in data affects the accuracy of the charts developed based on the iid assumption. One way to reduce the impact of autocorrelation is to estimate the value of the observation from a model and compute the error between the measured and estimated values. The errors, also called residuals, are assumed to have a Normal distribution with zero mean. Consequently regular SPM charts such as Shewhart or CUSUM charts could be used on the residuals to monitor process behavior. This method relies on the existence of a process model that can predict the observations at each sampling time. Various techniques for empirical model development are presented in Chapter 4. The most popular modeling technique for SPM has been time series models [1, 202] outlined in Section 4.4, because they have been used extensively in the statistics community, but in reality any dynamic model could be used to estimate the observations. If a good process model is available, the prediction errors (residual) e k) = y k)—y k) can be used to monitor the process status. If the model provides accurate predictions, the residuals have a Normal distribution and are independently distributed with mean zero and constant variance (equal to the prediction error variance). 

Si is the covariation matrix, and rq is the sample size of class i. Thus, based on these equations we can calculate the coefficients of the LDF and the threshold constant c using the values of characteristics of site and non-site sequences from the training sets and then test the accuracy of the LDF on the test set data. The significance of a given characteristic or set of characteristics can be estimated by the generalized distance between two classes (called the Mahalonobis distance or D2) ... 

During the estimation of parameters the test value Tjd) (41) will be calculated for each point Xu i). In case more points with a significant error will be detected, only the point with the largest test value will be eliminated. Subsequently a new estimation of the parameters will be carried out and the point with the largest value will be eliminated again. This sucessive elimination — called iterative data-snooping — is repeated until the rest of points are compatible with the model. Based on the rest of points the final parameters with their accuracy will be determined. 

---