Probabilistic scoring

Murvai, J., K. Vlahovicek, and S. Pongor, A simple probabilistic scoring method for protein domain identification. Bioinformatics, 2000. 16(12) p. 1155-6. 

Wilson, C. A., J. Kreychman, and M. Gerstein. 2000. Assessing annotation transfer for genomics Quantifying the relations between protein sequence, structure and function through traditional and probabilistic scores. J Mol Biol 297 233 49. 

Candidate peptide sequences from the mixture are later identified using a computerized database search algorithm (e.g., SEQUEST) [14] and validated using the STATQUEST probabilistic scoring program [15]. 

In the probabilistic scoring approach, a scoring profile is defined to reflect the profile of properties required for an ideal compound in the context of a project (an example is shown in Figure 8.11). This profile may include simple calculated characteristics, predicted properties, or experimental endpoints. Underlying each of the property criteria is a desirability function that defines the importance of the property to the overall objective of the project and the acceptable compromises. These desirability functions are defined in terms of the impact of the property on the chance of success of the compound a low desirability indicates a low chance of success, or equivalently a high risk, due to the value of the property. Thus, the overall score will reflect the best estimate of the overall chance of success of a compound. As a probability, the overall score will be between zero and one and is multiplicative with respect to the contributions of the individual properties. 

An example application of probabilistic scoring is given in Section 8.4. 

Figure 15.3 This graph shows the scores for a set of 40 compounds, as calculated using the probabilistic scoring method [31]. The compounds are ordered along the x-axis from the highest to lowest scoring and the score is plotted on the y-axis. Error bars indicate the uncertainty (1 standard deviation) in the score...

Common methods of risk assessment include relative risk numerical scoring methods and numerical probabilistic methods. With relative methods, scores and weighting factors are assigned to various risk factors and grouping of factors, based on system attributes, and an overall risk score is generated. The methods determine a score for likelihood of failure, the severity of the consequences of failure, and the... 

The Joint Entropy-based Diversity Analysis (JEDA) is a method to select representative subsets of compounds from combinatorial libraries by using a scoring function based on the Shannon s entropy and implemented in a probabilistic search algorithm [Landon and Schaus, 2006]. 

To select the optimal subset of compounds, that is, a set of compounds having the maximal chemical diversity, a probabilistic search algorithm is applied, which consists in selecting a subset of compounds based on a probability assigned to each compound. This algorithm optimizes the joint entropy (/H) of the subset of selected compounds. The task is performed iteratively, assigning each ith compound an initial uniform probability Pi = 1 fn, then calculating the score S that is added to the previous compound probability as... 

P(r xj(r)) denotes the posterior probability for the true class of Xj, since the notation Xj(r) indicates that the true class of Xj is class r. Another simple probabilistic measure results when the appreciation score is... 

---