mas5calls                package:affy                R Documentation

_M_A_S _5._0 _A_b_s_o_l_u_t_e _D_e_t_e_c_t_i_o_n

_D_e_s_c_r_i_p_t_i_o_n:

     Performs the Wilcoxon signed rank-based gene expression
     presence/absence  detection algorithm first implemented in the
     Affymetrix Microarray Suite version 5.

_U_s_a_g_e:

     mas5calls(object,...)

     mas5calls.AffyBatch(object, ids = NULL, verbose = TRUE, tau = 0.015,
                         alpha1 = 0.04, alpha2 = 0.06,
                         ignore.saturated=TRUE) 

     mas5calls.ProbeSet(object, tau = 0.015, alpha1 = 0.04, alpha2 = 0.06,
                        ignore.saturated=TRUE) 

     mas5.detection(mat, tau = 0.015, alpha1 = 0.04, alpha2 = 0.06,
                    exact.pvals = FALSE, cont.correct = FALSE)

_A_r_g_u_m_e_n_t_s:

  object: An object of class 'AffyBatch' or 'ProbeSet'

     ids: probeset IDs for which you want to compute calls

     mat: an n-by-2 matrix of paired values (pairs in rows), PMs first
          col

 verbose: logical. It 'TRUE' status of processing is reported

     tau: a small positive constant 

  alpha1: a significance threshold in (0,alph2)

  alpha2: a significance threshold in (alpha1,0.5) 

exact.pvals: a boolean controlling whether exact p-values are computed
          (irrelevant if n<50 and there are no ties).  Otherwise the
          normal approximation is used

ignore.saturated: if true do the saturation correction described in the
          paper, with a saturation level of 46000 

cont.correct: a boolean controlling whether continuity correction is
          used in the p-value normal approximation  

     ...: any of the above arguments that applies

_D_e_t_a_i_l_s:

     This function performs the hypothesis test:

     H0: median(Ri) = tau, corresponding to absence of transcript H1:
     median(Ri) > tau, corresponding to presence of transcript

     where Ri = (PMi - MMi) / (PMi + MMi) for each i a probe-pair in
     the  probe-set represented by data.

     Currently exact.pvals=TRUE is not supported, and cont.correct=TRUE
     works but does not give great results (so both should be left as
     FALSE).  The defaults for tau, alpha1 and alpha2 correspond to
     those in MAS5.0.

     The p-value that is returned estimates the usual quantity:

     Pr(observing a more "present looking" probe-set than data | data
     is absent)

     So that small p-values imply presence while large ones imply
     absence of  transcript.  The detection call is computed by
     thresholding the p-value as  in:

     call "P" if p-value < alpha1 call "M" if alpha1 <= p-value <
     alpha2 call "A" if alpha2 <= p-value

     This implementation has been validated against the original MAS5.0
     implementation with the following results (for exact.pvals and
     cont.correct set to F):

     Average Relative Change from MAS5.0 p-values:38% Proportion of
     calls different to MAS5.0 calls:1.0%

     where "average/proportion" means over all probe-sets and arrays,
     where  the data came from 11 bacterial control probe-sets
     spiked-in over a range of concentrations (from 0 to 150 pico-mols)
     over 26 arrays.  These are the spike-in data from the GeneLogic
     Concentration Series Spikein Dataset.

     Clearly the p-values computed here differ from those computed by
     MAS5.0 - this will be improved in subsequent releases of the affy
     package.   However the p-value discrepancies are small enough to
     result in the call being very closely aligned with those of MAS5.0
     (99 percent were identical on the validation set) - so this
     implementation will still be of use.

     The function 'mas5.detect' is no longer the engine function for
     the others. C code is no available that computes the wilcox test
     faster. THe function is kept so that people can look at the R code
     (instead of C)

_V_a_l_u_e:

     'mas5.detect' returns a list containing the following components: 

    pval: a real p-value in [0,1] equal to the probability of observing
          probe-level intensities that are more present looking than
          data assuming the data represents an absent transcript; that
          is a transcript is more likely to be present for p-values
          closer 0.

    call: either "P", "M" or "A" representing a call of present,
          marginal or absent; computed by simply thresholding pval
          using alpha1 and


     The 'mas5calls' method for 'AffyBatch' returns an 'ExpressionSet'
     with calls accessible with 'exprs(obj)' and p-values available
     with 'assayData(obj)[["se.exprs"]]'. The code{mas5calls} for
     'ProbeSet' returns a list with vectors of calls and pvalues.

_A_u_t_h_o_r(_s):

     Crispin Miller, Benjamin I. P. Rubinstein, Rafael A. Irizarry

_R_e_f_e_r_e_n_c_e_s:

     Liu, W. M. and Mei, R. and Di, X. and Ryder, T. B. and Hubbell, E.
     and Dee, S. and Webster, T. A. and Harrington, C. A. and Ho, M. H.
     and Baid, J. and Smeekens, S. P. (2002) Analysis of high density
     expression microarrays with signed-rank call algorithms,
     Bioinformatics, 18(12), pp. 1593-1599.

     Liu, W. and Mei, R. and Bartell, D. M. and Di, X. and Webster, T.
     A. and Ryder, T. (2001) Rank-based algorithms for analysis of
     microarrays, Proceedings of SPIE, Microarrays: Optical
     Technologies and Informatics, 4266.

     Affymetrix (2002) Statistical Algorithms Description Document,
     Affymetrix Inc., Santa Clara, CA, whitepaper. <URL:
     http://www.affymetrix.com/support/technical/whitepapers/sadd_whitepaper.pdf>,
     <URL:
     http://www.affymetrix.com/support/technical/whitepapers/sadd_whitepaper.pdf>

_E_x_a_m_p_l_e_s:

     data(affybatch.example)
     PACalls <- mas5calls(affybatch.example)

