CHANGES IN VERSION 1.7.1
------------------------

BUG FIXED

    o Null eigenvectors for TBA with null variance or NA values in the training
      set are returned correctly


CHANGES IN VERSION 1.3.7
------------------------

UPDATED VIGNETTE

    o Package vignette has been updated to address the new features

ALSO MODELS' P-VALUES AND RHO ARE RETURNED IN CROSS-VALIDATION

    o For each fold, for each gene for which a GReX could be imputed, the
      following values are now returned
        
        o Squared correlation of the model's predictions with training data

        o P-value of the model

        o Corrected p-value of the model with benjamini-hochberg procedure

        o Correlation of the model's predictions with validation data

        o Squared correlation of the model's predictions with validation data

        o P-value of the correlation test of the model's predictions with
          validation data


CHANGES IN VERSION 1.3.5
------------------------

P-VALUE OF THE CORRELATION TEST

    o P-value of the cor.test() between predicted GReX and real expression
      values of genes is returned in when performing cross-validation

    o When using affiXcanTrain in cross-validation mode, three main values
      for each gene are therefore returned: rho and rho squared (see changes
      in v 1.3.1), and the p-value of cor.test()


CHANGES IN VERSION 1.3.3
------------------------

UPDATED DOCUMENTATION

    o Formatting of functions documentation has been improved

    o Important: vignette is still outdated (AffiXcan 1.2.0)


CHANGES IN VERSION 1.3.2
------------------------

POPULATION STRUCTURE COVARIATES ARE OPTIONAL

    o Providing population structure covariates is now not mandatory to perform
      models training


CHANGES IN VERSION 1.3.1
------------------------

K-FOLD CROSS-VALIDATION IS SUPPORTED

    o ANOVA p-value < 0.05 to assess prediction significance is not used
      anymore; instead:
    
    o A k-fold cross-validation on the training dataset may be performed;
      k can be defined by the user
    
    o Pearson correlation coefficients (R) and determination coefficients (R^2)
      between predicted GReX and real expression values of genes are computed
    
    o In literature, GReX of genes for which the mean of the R^2 is above 0.01
      are considered non-randomly predicted, according to the new benchmarking
      standards [ref] 


CHANGES IN VERSION 1.1.3
------------------------

BUG FIXES

    o Fix a bug in generating the vignette for 'devel' version of Bioconductor