timedreamer / ricetfs Goto Github PK

In the paper, it says:

After sequencing, 185, 6125, 220, 455 peaks were finally obtained respectively in OX127H, OX127N, OX129H and OX129N.

However, I got: 52--> NAC127H; 179--> NAC127N; 132-->NAC129H; 139-->NAC129N.

I tried some several settings in masc2, but always have fewer peaks than reported.
The method that got more peaks would be analyzing three reps individually with the control. Then combine all the peaks.

For GSE72186 OsNF-YB1 ChIP, genes that found in either replicate were kept.
I think this is similar to what the paper did:

Overlapping peaks (having a stronger but narrower peak in at least one replicate) between two biological replicates were used for further analysis.

The final file is the GSE72186_rep12.MACS2_summits_gene_list_2404

It's not clear how they call peaks for this TF in the paper. I have many issues with the dataset.

1. There are no control or input for Roc10-myc α-myc.
2. The GEO supplementary tables are very confusing. What's the meaning for 46, H3K4me3, H3K27me3? 46 probably means transgenic plants?
3. How were the binding sites in Table 1 calculated? I thought they were anti-myc. But in the GEO tables, many of them were labeled something else, like chr03-7132344-7133928-NT-H3K4me3-7455. Besides, the peak range is too long, probably from --broad in MACS? But should TF use this setting?

In summary, I re-did the analysis using the same pipeline as PE.
Only SRR5164450 Roc10-myc α-myc ChIPseq was used for peak calling.
I have to use --nomodel --extsize 147, otherwise, no peaks could be found.

In the end I got 161 peaks correspondent to 85 genes (within 2kb region)

1. The data was from this paper.

2. In the paper TableS3, there are 14 libraries. However, only 12 in the GEO/SRA. Missing one WT-48HAF and one WT-72-HAF.

3. There is only one replicate of genotype x time combination. For example, one replicate of OE78 x 48HAF. So there is no way to do good the statistical tests. See table below.
   

4. The conclusion: give up MADS78/79. Unless we want to combine 48HAF and 72HAF and use as replicates.

Does one TF allow multiple DE gene list?
For example, there are two lines fo mutants bzip76-1 and bzip76-2 with wildtype.
Do I keep two lists as bzip76-1 vs wt and bzip76-2 vs wt?

Or combine them together? Is so, should I use a union or intersection?

This paper has a root specific OX line, whole body OX line and a KO mutant, but I could only find one replicate for each of the mutant lines - so we probably can't use this right? - I think it is fine to skip since we already have OsNAC6 from the paper with the 4 NACs.

1. there is no treatment info from the SRP meta-data. So I used the GSM ID to find the treatment info from GSE.
   

2. I added the info to the treatment column in SraRunTable_20200430_11SRP.tsv file.

3. There are two SRA correspondent to one GSM. At first, I thought they split the single-end files because of the file size. However, they all have the same number of reads. So I think they are actually paired-end reads! But which is read1 and which is read2 is not clear.

In conclusion, unless we really want this TF. I can try to reallign. Otherwise, I gave up this TF.