Design matrix for ANOVA using limma

Question

[gustavofw]

Hi,

I am new to bioinformatics, coding, R and my stats knowledge is far from ideal. I need to compare three groups and it seems I can do that using limma since lmFit + eBayes calculates F values which are also used by ANOVA, right? To my understating, I would have to make sure my design matrix is appropriate and use contrast.fit() to make the group comparisons; however, I have no idea where to start. I searched for ANOVA in limma's manual but very little explanation was provided (or I was not able to find it). I would appreciate it if someone could help me design the matrix and how to perform the ANOVA. If possible, I would also appreciate a good resource where I can learn to design theses matrices.

I have 3 groups (G), each with 2 biological replicates (B) and 3 technical replicates (R).

| G1_B1_R1 | G1_B1_R2 | G1_B1_R3 | G1_B2_R1 | G1_B2_R2 | G1_B2_R3 |

| G2_B1_R1 | G2_B1_R2 | G2_B1_R3 | G2_B2_R1 | G2_B2_R2 | G2_B2_R3 |

| G3_B1_R1 | G3_B1_R2 | G3_B1_R3 | G3_B2_R1 | G3_B2_R2 | G3_B2_R3 |

Please, let me know if you need any extra information or if it is too confusing and I will do my best to make it as clear as possible.

Cheers, Luiz

Answer 1

[h.mon]

The limma Users Guide is a great resource for learning design matrices for microarrays experiments (you can add to the list the edgeR Users Guide and the "Analyzing RNA-seq data with DESeq2" document). The limma users Guide explains how to set up the design and the contrasts, explains how to perform significance tests, and has plenty of examples with several experimental designs. But you will have to read the guide carefully from start to end, searching for "anova" and hoping to find a solution won't do it.

Regarding the technical replicates, limma users Guide suggestion is to use duplicateCorrelation() to take them into account. You tagged your question with proteomics, is this an protein array experiment? How good is technical replicability for protein arrays?