Hello!
I have rna seq data (2 samples) processed using cufflinks-cuffdiff-cummerbund. Now i want to carry out rt-qpcr experiment to confirm the results of DE-genes, and i need reference genes that are stable in both my samples. I suppose that stable genes are those where p value is < 0.05 (significant) and log2 fold change tends to zero. So:
test_id gene_id gene locus sample_1 sample_2 status value_1 value_2 log2(fold_change) test_stat p_value q_value significant
XLOC_000004 XLOC_000004 gyrB NZ_CP046257.1:5895-7959 37DG 37NG OK 13.6418 10.7901 -0.338319 -1.655 0.02635 0.0680931 yes # stable gene, suitable to be reference
XLOC_000005 XLOC_000005 gyrA NZ_CP046257.1:8003-10493 37DG 37NG OK 7.80287 8.31905 0.0924146 0.407881 0.5788 0.692229 no # not, because p value is high, the observation is not reliable, high probability of randomness
XLOC_000689 GKZ92_RS07450,sfnG 143.727 16.7024 -3.10521 -17.4389 5e-05 0.000410086 yes # DE-expressed gene
am i right or wrong?
Thank you.
0 answers
No answers yet.
Log in to answer this question.
The pvalue tests the fold change is zero, hence having pvalue < 0.05 is not a good choice for "non-changing genes". DESeq2 has a test than allows for testing against differential expression (altHypothesis lessAbs). For your purpose it might be enough though to check the common suspects such as actin, Gapdh or VCP, just checking the normalized counts by eye, and then look at the pvalue and logFC. logFC should be close to zero, gene should be strongly expressed and pvalue should be high. Together with the knowledge that these genes are usually used as reference that might be sufficient here. I anyway do not think that qPCR is a meaningful way of validation as you need many more replicates to get statistical significances with tests such as the Wilcox, and you need many primer pairs to get somewhat a feel for the npise of your setup. We almost never do it these days in our lab, rather seek for experimental validation whether the RNA-seq significant genes are biologically meaningful. You any way have to do that I guess.
So, this gene
XLOC_000005 XLOC_000005 gyrA NZ_CP046257.1:8003-10493 37DG 37NG OK 7.80287 8.31905 0.0924146 0.407881 0.5788 0.692229 no
is good choise, isnt it? sorry, maybe i didnt understand you correctly. i'm trying to find the logic how using only cuffdiff output we can realise that the gene is reliable to be reference
Here are the genes commonly used as references in rt-qpcr experiments.
test_id gene_id gene locus sample_1 sample_2 status value_1 value_2 log2(fold_change) test_stat p_value q_value significant
XLOC_000004 XLOC_000004 gyrB NZ_CP046257.1:5895-7959 37DG 37NG OK 13.6418 10.7901 -0.338319 -1.655 0.02635 0.0680931 yes
XLOC_000005 XLOC_000005 gyrA NZ_CP046257.1:8003-10493 37DG 37NG OK 7.80287 8.31905 0.0924146 0.407881 0.5788 0.692229 no
XLOC_002953 XLOC_002953 ftsZ NZ_CP046257.1:3201366-3202542 37DG 37NG OK 7.53876 10.2517 0.443462 1.32097 0.07745 0.153534 no
XLOC_000001 XLOC_000001 dnaA NZ_CP046257.1:167-1700 37DG 37NG OK 7.86392 7.51386 -0.0656945 -0.218239 0.7576 0.830145 no
XLOC_003256 XLOC_003256 secA NZ_CP046257.1:3833300-3836120 37DG 37NG OK 11.3506 8.32 -0.448107 -2.34055 0.0022 0.0101551 yes
XLOC_003524 XLOC_003524 dnaJ NZ_CP046257.1:4463579-4464776 37DG 37NG OK 12.8647 10.909 -0.237909 -0.839309 0.2571 0.380424 no
XLOC_003678 XLOC_003678 dnaB NZ_CP046257.1:4902431-4904684 37DG 37NG OK 13.1982 8.5888 -0.61981 -2.93567 0.0001 0.000765932 yes
XLOC_000001 XLOC_000001 dnaA NZ_CP046257.1:167-1700 37DG 37NG OK 7.86392 7.51386 -0.0656945 -0.218239 0.7576 0.830145 no
XLOC_002938 XLOC_002938 dnaE NZ_CP046257.1:3178512-3182055 37DG 37NG OK 7.83071 7.94069 0.0201214 0.106732 0.8887 0.925003 no
XLOC_003044 XLOC_003044 GKZ92_RS15355,GKZ92_RS15360,dnaJ NZ_CP046257.1:3391213-3393917 37DG 37NG OK 38.8768 29.3521 -0.405448 -1.94025 0.011 0.0360566 yes
XLOC_002801 XLOC_002801 recA NZ_CP046257.1:2926713-2927757 37DG 37NG OK 17.0722 8.96362 -0.929495 -3.02203 0.0001 0.000765932 yes
XLOC_002709 XLOC_002709 gap NZ_CP046257.1:2727882-2728902 37DG 37NG OK 36.0558 7.99475 -2.17311 -7.44103 5e-05 0.000410086 yes
XLOC_003331 XLOC_003331 tuf NZ_CP046257.1:4012549-4013740 37DG 37NG OK 50.7987 20.8954 -1.28161 -7.31742 5e-05 0.000410086 yes
XLOC_003332 XLOC_003332 fusA NZ_CP046257.1:4013853-4015959 37DG 37NG OK 43.6613 13.7552 -1.66638 -11.0499 5e-05 0.000410086 yes
XLOC_000467 XLOC_000467 GKZ92_RS05135 NZ_CP046257.1:1119125-1122614 37DG 37NG OK 159.995 15.2328 -3.39278 -33.7141 5e-05 0.000410086 yes # rpoB
XLOC_000468 XLOC_000468 GKZ92_RS05140 NZ_CP046257.1:1122652-1126609 37DG 37NG OK 74.4598 13.2157 -2.49421 -23.5577 5e-05 0.000410086 yes # rpoB dna directed
XLOC_002725 XLOC_002725 GKZ92_RS12365 NZ_CP046257.1:2753383-2753683 37DG 37NG OK 175.286 34.1589 -2.35938 -5.92406 5e-05 0.000410086 yes # rpoW dna directed
XLOC_003271 XLOC_003271 GKZ92_RS17640 NZ_CP046257.1:3861306-3862359 37DG 37NG OK 52.6204 11.7296 -2.16546 -9.09525 5e-05 0.000410086 yes # rpoA
The logic would be to use stronglyexpressed genes with minimal or zero logFC and large pvalue. There is (to my knowledge, not a cuffdiff user) no test in cuffdiff against differential expression. I would not even use cuffdiff these days, it is old and superseded by newer and better methods. Use DESeq2 or edgeR.