FITS output¶
General¶
FITS infers population genetics parameters using the Approximate Bayesian Computation (ABC) method. The output of this method is a distribution of values that explain the observed allele frequencies with the highest probabilities (also called the posterior distribution). A common practice is to take the median of this distribution as the inferred value of the parameter under study.
The results below are outputted for all inferences.
Result header  Description 

median  The median value of the posterior distribution.

MAD  Median Absolute Deviation index (MAD) of the posterior distribution. 
min  The minimum value in the posterior distribution. 
max  the maximum value in the posterior distribution. 
pval  The result of a Levene’s test,
comparing between the prior and posterior distributions.
Successful inference process should yield a posterior distribution
that is significantly different from the prior.

Note
If the pvalue for Levene’s test is not significant (>=0.05) or Nµ is small (<1), FITS will mark the relevant line in the results with an asterisk (*). This result is considered unreliable.
Fitness inference results¶
In addition to the General reported values, in fitness inference more data are available:
Result header  Description 

allele  The allele for which the results are reported 
DEL(%)  The proportion of the posterior distribution with values below 1. 
NEU(%)  The proportion of the posterior distribution with values equal to 1. 
ADV(%)  The proportion of the posterior distribution with values above 1. 
category  A possible classification of the allele into {LETHAL,DEL,NEU,ADV}, based on the inferred fitness value. 
Note
Some fitness priors rarely choose the exact value of 1 and therefore NEU(%) will approach zero, even for neutral alleles.
Mutation rate inference results¶
FITS infers the mutation rates between all defined alleles. Accordingly, the output table contains the target allele in the first row and the source allele in the first column.
Note
In Evolve & Resequence (E&R) studies, when the population is homogeneous at first generation, in the absence of more information the inference of the rates between the minor allele to the major will be insignificant, so the pval should be taken into account.