BayesFold considers several types of evidence in ranking structures according to their probabilities. (For more information about how to view the evidences, see Section 1, “Showing and Hiding Tables or Columns”.)
A number of evidences are calculated as a summary of data from each individual sequence. For these, both Sequence and Structure tables display a raw score that shows the data for each sequence/structure combination (for example, the number of mismatches for each sequence folded into each structure). Both also display a best index (showing the sequence that is best for each structure and the structure that is best for each sequence) and an average of the raw score for each sequence and each structure (e.g. the average number of mismatches for a structure across all the sequences). In the structure table only, these evidences display a conditional probability showing how likely that evidence's data would be if the corresponding structure were the true one.
Folding Energy: the energy of a given sequence when folded into a given structure. Score is measured in kcal/mol. No separate conditional probability for Folding Energy is provided because pair probabilities are calculated directly from the energies.)
Mismatches: the number of mismatched pairs produced when a given sequence is threaded through a given structure. Score is an integer number of mismatches.
Other kinds of evidence are calculated from the whole sequence alignment. Consequently, each structure has a score, but there is no number associated with individual sequences. Therefore, these evidences appear only in the Structure table.
Mutual Information: the average of the difference between the information the alignment conveys about each pair of positions considered as a pair and considered independently. Score is measured in bits, with a range of 0 to 2 bits where 2 means high correlation.
Fraction Pairable: the probability that the actual fraction pairable data would be seen assuming that the associated structure is the correct one. Range is 0 to 1. Fraction pairable data are the fraction of sequences in the alignment in which a given pair of positions can make a valid pair; these data are also viewable on the drawing (see Section 4, “Displaying Data on the Structure Drawing”.)
Pair Probability: the probability that the actual pair probability data would be seen assuming that the associated structure is the correct one. Range is 0 to 1. Pair probability data are the average probability of each pair of positions across all sequences; these data are also viewable on the drawing (see Section 4, “Displaying Data on the Structure Drawing”.)
Chemical Mapping(s): chemical data indicating which bases are paired. Score is on a scale of 0-1 where 0 means no signal and 1 means maximum intensity. (In principle, the chemical mappings really do apply to individual sequences, but it's very rare for more than one sequence to be mapped so it's more useful to treat it as though it applied to the whole alignment.)
Although the data for all evidences are hidden by default, they can be displayed using each table's associated display palette (see Section 1, “Showing and Hiding Tables or Columns”.) The several types of data (best index, score, etc) associated with each evidence are discussed in more detail in Section 5, “Understanding the Evidence Columns”.