search

rhondabacher / Trendy

Trendy: Segmented regression analysis of expression dynamics for high-throughput ordered profiling experiments
https://bioconductor.org/packages/release/bioc/html/Trendy.html
6 stars 2 forks source link

Breakpoints are all NA #4

Open pablasimarjot opened 2 years ago

pablasimarjot commented 2 years ago

I am using normalized RNAseq counts as suggested in the manual to perform time-course analysis. But all the breakpoints are NA, even though the pvalues are significant. I used the following settings in the function (I tried with replicates and no replicates, but I get the same problem of NAs): test (Normalized counts) testvector (1,2,3,4)

result<- trendy(Data = test, tVectIn = testvector, minNumInSeg=3, saveObject = TRUE)

$Segment.Pvalues [,1] ABCF2 0.01305244 AASS 0.02047916 ABCB8 0.02812854 ABCC1 0.04653451 ABAT 0.05726043 ABCE1 0.07936356 AARS 0.08655334 AAMP 0.09254876 AAK1 0.09574053 AARSD1 0.10198953 ABCC4 0.10488407 ABCD3 0.11221271 AATF 0.11725859 AAAS 0.11954304 AAED1 0.12241864 AAGAB 0.14242967 ABCB6 0.14401778 ABCA2 0.15554329 ABCB10 0.15979219 AB019441.29 0.16654281 AAR2 0.16878536 AAMDC 0.16982965

$Breakpoints [,1] ABCF2 NA AASS NA ABCB8 NA ABCC1 NA ABAT NA ABCE1 NA AARS NA AAMP NA AAK1 NA AARSD1 NA ABCC4 NA ABCD3 NA AATF NA AAAS NA AAED1 NA AAGAB NA ABCB6 NA ABCA2 NA ABCB10 NA AB019441.29 NA AAR2 NA AAMDC NA

rhondabacher commented 2 years ago

In this case, trendy is detecting that there are no breakpoints. This is because minNumInSeg=3 requires 3 data points per segment and even 1 breakpoint would require 2 segments, each having 3 data-points. Since you only have 4 data-points, this is why it's returning no break-points. The p-values indicate the expression is changing up or down over time, linearly, no breakpoints. You can try setting minNumInSeg=2, but be very cautious for false positives. A different method may be more appropriate for this shorter time course.