This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Unusual Cox Proportional Hazards Results for Highly Expressed Microarray Probes

I did a survival analysis in two ways. Firstly, I dichotomised the survival data into two groups.

  1. Dead from cancer within one year.
  2. Alive more than four years, with no sign of relapse.

Then, I used limma to build a linear model and find differentially expressed probes.

Secondly, I used Cox proportional hazards regression. I fitted one model per probe.

hazardModels <- lapply(1:nrow(expression), function(probe) coxph(survivalData ~ expression[probe, ]))

I plotted the coefficients from the two methods, to check concordance. Most probes have similar coefficients, and the scatterplot is quite linear. I plotted raw expression values for some probes that were in disagreement. It is surprising that the Cox proportional hazards model is detecting many genes that have no probe expression difference between groups, but are highly expressed. These probes are not detected by the first method. What is the statistical explanation for this ?

microarray survival proportional hazards
• 3,771 views • link
updated by Ram Ram • written by dario.garvan dario.garvan

It'd be rather helpful to see an example.

• 2 views • link
written by Devon Ryan Devon Ryan 
expression <- c(15.3103311933149, 15.0157174552731, 15.135793474893,
14.9198927859697, 14.9897673751516, 15.3062827467706, 15.3103311933149,
15.2070521612507, 15.059556710088, 15.135793474893, 15.1313866460974,
14.9813325911164, 15.0831610844229, 15.0781787120161, 15.2609923941384,
14.7705795759363, 15.3384294956832, 14.8152507315718, 15.3012062540704,
15.2753063785538, 15.1954229268207, 15.3103311933149, 15.3062827467706,
15.3750848737185, 15.1578260359732, 15.321173185739, 15.2609923941384,
14.9149292674096, 14.9377815644205, 15.0874554648443, 15.1263951394799,
15.1912205682114, 14.0848402429873, 15.3674426187413, 14.7158303376285,
14.8152507315718, 15.1400173563972, 15.4431428726152, 15.2466239892204,
14.9691791905281, 15.3103311933149, 15.3160174597845, 15.1093785260608,
15.2697245998598, 14.8790422727129, 14.7360317386087, 15.3103311933149,
15.0296043828935, 14.7158303376285, 15.3750848737185, 15.0636287648631,
15.2070521612507, 15.1504088368176, 15.0831610844229, 15.3897113254823,
15.1954229268207, 15.0296043828935, 15.1263951394799, 15.1720391013973,
14.9339393180207, 15.221579147451, 15.2855669160713, 15.0874554648443,
15.2171597176676, 15.1175343571932, 15.2559796011832, 15.11361722107,
15.3750848737185)

days <- c(3528, 3509, 3289, 443, 1905, 658, 2939, 507, 238,
1326, 1888, 53, 362, 275, 2151, 256, 70, 296, 4628,
2204, 3357, 490, 1913, 2974, 3769, 27, 169, 4357, 43,
2969, 3044, 2885, 226, 2408, 1748, 714, 215, 67, 1200,
377, 3921, 1499, 2115, 176, 3650, 233, 896, 454, 455,
189, 2893, 2932, 241, 193, 1251, 752, 1662, 344, 3298,
2948, 326, 2809, 235, 1570, 201, 1888, 671, 575)

status <- c(FALSE, FALSE, FALSE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE,
TRUE, TRUE, TRUE, TRUE, FALSE, TRUE, TRUE, TRUE, FALSE, FALSE,
FALSE, TRUE, TRUE, FALSE, FALSE, TRUE, TRUE, FALSE, TRUE, FALSE,
FALSE, FALSE, TRUE, TRUE, FALSE, TRUE, TRUE, TRUE, FALSE, TRUE,
FALSE, FALSE, FALSE, TRUE, FALSE, TRUE, TRUE, TRUE, TRUE, TRUE,
FALSE, FALSE, TRUE, TRUE, FALSE, TRUE, TRUE, TRUE, FALSE, FALSE,
TRUE, FALSE, TRUE, FALSE, TRUE, FALSE, TRUE, TRUE)

library(survival)
survivalData <- Surv(days, status)
coxph(survivalData ~ expression)

library(ggplot2)
plotData <- data.frame(years = days / 365,
                       expression = expression,
                       status = status)
ggplot(plotData, aes(x = years, y = expression, colour = status)) + ylim(4, 16) + geom_point()
• 2 views • link
updated by Ram Ram • written by dario.garvan dario.garvan

do you get similar results for different probes from the same gene?

are there outliers for expression[probe, ] that may adversely affect the survival analysis?

have you log-transformed, or similar, expression[probe, ] and do your histograms for the same look 'normal'?

• 2 views • link
written by russhh russhh

1 answer

The explanation is that there is a small change in the ratio of events to survivors as the expression level increases. The trend is exaggerated by the reduced variability of the highly expressed genes being near the maximum limit of measurement.

• 1 views • link
written by dario.garvan dario.garvan

Log in to answer this question.

More posts like this