In averissimo/glmSparseNetPaper:

Robustez do método - Runs dá coisas diferentes - Uma forma para escolher um bom modelo para treinar (simulando várias partições) - Em particular nós usamos esta framework para escolher um modelo particular para mostrar resultados - Isto não é visto em resultados de papers-- pq há muitos runs q não dão significativos

Rank dos 3 métodos

O que usar como tempo de follow-up (inconsistência entre melanoma e breast)

Dar nomes aos resultados - estabilidade - consistência de genes seleccionados

Identificar as dificuldades em avaliar os resultados

devtools::load_all()

n.cores <- 14

my.render('Analysis.Rmd', output_file = 'brca-center-1000-random.html', 
          params = list(normalization = 'center', 
                        project       = 'brca.data.2018.09.13',
                        tissue        = 'primary.solid.tumor',
                        subtract.surv.column = NULL,
                        balanced.sets = FALSE,
                        ntimes        = 1000,
                        n.cores       = n.cores,
                        target.vars   = list(glmnet = list(alpha = .2, vars = 39), 
                                             hub    = list(alpha = .2, vars = 56), 
                                             orphan = list(alpha = .2, vars = 42))
                        ))

my.render('Analysis.Rmd', output_file = 'skcm-center-1000-random.html', 
          params = list(normalization = 'center', 
                        project       = 'skcm.data.2018.09.11', 
                        tissue        = 'metastatic',
                        balanced.sets = FALSE,
                        n.cores       = n.cores,
                        subtract.surv.column = 'days_to_submitted_specimen_dx',
                        ntimes        = 1000, 
                        target.vars   = list(glmnet = list(alpha = .1, vars = 116), 
                                             hub    = list(alpha = .1, vars = 41),
                                             orphan = list(alpha = .1, vars = 111))
                        ))

n.cores <- 14
ntimes <- 5000

#
# Completely random training/test sets instead of balanced
# 

my.render('Analysis.Rmd', output_file = sprintf('brca-center-%d.html', ntimes), 
          params = list(normalization = 'center', 
                        project       = 'brca.data.2018.09.13',
                        tissue        = 'primary.solid.tumor',
                        subtract.surv.column = NULL,
                        balanced.sets = TRUE,
                        ntimes        = ntimes,
                        n.cores       = n.cores,
                        target.vars   = list(glmnet = list(alpha = .2, vars = 39), 
                                             hub    = list(alpha = .2, vars = 56), 
                                             orphan = list(alpha = .2, vars = 42))
                        ))

my.render('Analysis.Rmd', output_file = sprintf('skcm-center-%d.html', ntimes), 
          params = list(normalization = 'center', 
                        project       = 'skcm.data.2018.09.11', 
                        tissue        = 'metastatic',
                        balanced.sets = TRUE,
                        subtract.surv.column = 'days_to_submitted_specimen_dx',
                        ntimes        = ntimes, 
                        n.cores       = n.cores,
                        target.vars   = list(glmnet = list(alpha = .1, vars = 116), 
                                             hub    = list(alpha = .1, vars = 41),
                                             orphan = list(alpha = .1, vars = 111))
                        ))

my.render('Analysis.Rmd', output_file = sprintf('kirc-center-%d.html', ntimes), 
          params = list(normalization = 'center', 
                        project       = 'kirc.data.2018.10.17', 
                        tissue        = 'primary.solid.tumor',
                        balanced.sets = TRUE,
                        subtract.surv.column = NULL,
                        ntimes        = ntimes, 
                        n.cores       = n.cores,
                        target.vars   = list(glmnet = list(alpha = .2, vars = 69), 
                                             hub    = list(alpha = .1, vars = 41),
                                             orphan = list(alpha = .1, vars = 38))
                        ))

my.render('Analysis.Rmd', output_file = sprintf('lgg-center-%d.html', ntimes), 
          params = list(normalization = 'center', 
                        project       = 'lgg.data.2018.10.17', 
                        tissue        = 'primary.solid.tumor',
                        balanced.sets = TRUE,
                        subtract.surv.column = NULL,
                        ntimes        = ntimes, 
                        n.cores       = n.cores,
                        target.vars   = list(glmnet = list(alpha = .1, vars = 175), 
                                             hub    = list(alpha = .2, vars = 51),
                                             orphan = list(alpha = .2, vars = 101))
                        ))

my.render('Analysis.Rmd', output_file = sprintf('luad-center-%d.html', ntimes), 
          params = list(normalization = 'center', 
                        project       = 'luad.data.2018.10.17', 
                        tissue        = 'primary.solid.tumor',
                        balanced.sets = TRUE,
                        subtract.surv.column = NULL,
                        ntimes        = ntimes, 
                        n.cores       = n.cores,
                        target.vars   = list(glmnet = list(alpha = .1, vars = 110), 
                                             hub    = list(alpha = .1, vars = 56),
                                             orphan = list(alpha = .1, vars = 97))
                        ))

my.render('Analysis.Rmd', output_file = sprintf('prad-center-%d.html', ntimes), 
          params = list(normalization = 'center', 
                        project       = 'prad.data.2018.10.11', 
                        tissue        = 'primary.solid.tumor',
                        balanced.sets = TRUE,
                        subtract.surv.column = NULL,
                        ntimes        = ntimes, 
                        n.cores       = n.cores,
                        target.vars   = list(glmnet = list(alpha = .6, vars = 3), 
                                             hub    = list(alpha = .1, vars = 13),
                                             orphan = list(alpha = .1, vars = 5))
                        ))

my.render('Analysis.Rmd', output_file = sprintf('ucec-center-%d.html', ntimes), 
          params = list(normalization = 'center', 
                        project       = 'ucec.data.2018.10.17', 
                        tissue        = 'primary.solid.tumor',
                        balanced.sets = TRUE,
                        subtract.surv.column = NULL,
                        ntimes        = ntimes, 
                        n.cores       = n.cores,
                        target.vars   = list(glmnet = list(alpha = .1, vars = 170), 
                                             hub    = list(alpha = .1, vars = 38),
                                             orphan = list(alpha = .1, vars = 125))
                        ))


my.render('Analysis.Rmd', output_file = sprintf('ov-center-%d.html', ntimes), 
          params = list(normalization = 'center', 
                        project       = 'ov.data.2018.11.30',
                        tissue        = 'primary.solid.tumor',
                        subtract.surv.column = NULL,
                        balanced.sets = TRUE,
                        ntimes        = ntimes,
                        n.cores       = n.cores,
                        target.vars   = list(glmnet = list(alpha = .2, vars = 30), 
                                             hub    = list(alpha = .2, vars = 55), 
                                             orphan = list(alpha = .2, vars = 110))
                        ))

n.cores <- 10
seed.additional <- 1


my.render('OptimalVariableSize.Rmd', 
          output_file = sprintf('skcm-optimal-%d.html', seed.additional + 1),
          params = list(seed.additional = seed.additional,
                        project         = 'skcm.data.2018.09.11',
                        subtract.surv.column = 'days_to_submitted_specimen_dx',
                        n.cores         = n.cores,
                        tissue          = 'metastatic'))


my.render('OptimalVariableSize.Rmd', 
          output_file = sprintf('brca-optimal-%d.html', seed.additional + 1), 
          params = list(seed.additional = seed.additional,
                        project         = 'brca.data.2018.09.13',
                        n.cores         = n.cores,
                        tissue          = 'primary.solid.tumor'))

my.render('OptimalVariableSize.Rmd', 
          output_file = sprintf('prad-optimal-%d.html', seed.additional + 1),
          params = list(seed.additional = seed.additional,
                        project         = 'prad.data.2018.10.11',
                        n.cores         = n.cores,
                        tissue          = 'primary.solid.tumor'))

my.render('OptimalVariableSize.Rmd', 
          output_file = sprintf('ucec-optimal-%d.html', seed.additional + 1),
          params = list(seed.additional = seed.additional,
                        project         = 'ucec.data.2018.10.17',
                        n.cores         = n.cores,
                        tissue          = 'primary.solid.tumor'))

my.render('OptimalVariableSize.Rmd', 
          output_file = sprintf('kirc-optimal-%d.html', seed.additional + 1), 
          params = list(seed.additional = seed.additional,
                        project         = 'kirc.data.2018.10.17',
                        n.cores         = n.cores,
                        tissue          = 'primary.solid.tumor'))

my.render('OptimalVariableSize.Rmd', 
          output_file = sprintf('luad-optimal-%d.html', seed.additional + 1), 
          params = list(seed.additional = seed.additional,
                        project         = 'luad.data.2018.10.17',
                        n.cores         = n.cores,
                        tissue          = 'primary.solid.tumor'))

my.render('OptimalVariableSize.Rmd', 
          output_file = sprintf('lgg-optimal-%d.html', seed.additional + 1), 
          params = list(seed.additional = seed.additional,
                        project         = 'lgg.data.2018.10.17',
                        n.cores         = n.cores,
                        tissue          = 'primary.solid.tumor'))

averissimo/glmSparseNetPaper documentation built on Jan. 25, 2021, 12:11 p.m.