a question about Saturation analysis in MEDIPS

[lei]

hi,
I am now using MEDIPS for MCIP-seq analysis.
For the Saturation analysis, I don't know how to explain the results. So which value I should focuse in order to answer the question that how many reads is sufficient for further analysises?
for example, here are the results for two samples, how to know whether the reads number is enough or not.

sample 1:

$distinctSets
[,1] [,2]
[1,] 0 0.0000000
[2,] 421516 0.4890383
[3,] 843032 0.6566207
[4,] 1264548 0.7414630
[5,] 1686064 0.7923945
[6,] 2107580 0.8271925
[7,] 2529096 0.8514569
[8,] 2950612 0.8699040
[9,] 3372128 0.8844299
[10,] 3793644 0.8959736
[11,] 4215166 0.9053887

$estimation
[,1] [,2]
[1,] 0 0.0000000
[2,] 421516 0.4972269
[3,] 843032 0.6631622
[4,] 1264548 0.7467590
[5,] 1686064 0.7974522
[6,] 2107580 0.8311024
[7,] 2529096 0.8552331
[8,] 2950612 0.8731673
[9,] 3372128 0.8873165
[10,] 3793644 0.8985876
[11,] 4215160 0.9077852
[12,] 4636676 0.9153314
[13,] 5058192 0.9218284
[14,] 5479708 0.9274455
[15,] 5901224 0.9323282
[16,] 6322740 0.9365573
[17,] 6744256 0.9402298
[18,] 7165772 0.9434950
[19,] 7587288 0.9465256
[20,] 8008804 0.9491727
[21,] 8430333 0.9516126

$numberReads
[1] 8430333

$maxEstCor
[1] 8.430333e+06 9.516126e-01

$maxTruCor
[1] 4.215166e+06 9.053887e-01

sample 2:

$distinctSets
[,1] [,2]
[1,] 0 0.0000000
[2,] 677985 0.6015521
[3,] 1355970 0.7518315
[4,] 2033955 0.8192142
[5,] 2711940 0.8577985
[6,] 3389925 0.8830463
[7,] 4067910 0.9005583
[8,] 4745895 0.9136279
[9,] 5423880 0.9235792
[10,] 6101865 0.9314847
[11,] 6779856 0.9379848

$estimation
[,1] [,2]
[1,] 0 0.0000000
[2,] 677985 0.6054022
[3,] 1355970 0.7542983
[4,] 2033955 0.8216820
[5,] 2711940 0.8597790
[6,] 3389925 0.8845496
[7,] 4067910 0.9017873
[8,] 4745895 0.9148084
[9,] 5423880 0.9246916
[10,] 6101865 0.9325375
[11,] 6779850 0.9387990
[12,] 7457835 0.9440903
[13,] 8135820 0.9485184
[14,] 8813805 0.9523514
[15,] 9491790 0.9555902
[16,] 10169775 0.9585080
[17,] 10847760 0.9609586
[18,] 11525745 0.9631440
[19,] 12203730 0.9651430
[20,] 12881715 0.9669121
[21,] 13559713 0.9684958

$numberReads
[1] 13559713

$maxEstCor
[1] 1.355971e+07 9.684958e-01

$maxTruCor
[1] 6.779856e+06 9.379848e-01

Thanks for your help!

[nilshomer]

Please do not cross post.

[NearyJL78]

EDIT - the paragraph below is referring to the next steps in MEDIPS analysis, sorry! I'm typing the answer to your question (AS I UNDERSTAND IT) after the ***


I am just starting with MEDIPS also, also looking at MeDIP-seq data. I am still somewhat puzzled by the analysis, but I know the higher the correlation values the better. Roughly, they represent how much your sample has been enriched for CpGs if you imagine the genome to have a value of 1. The higher above one you are, the more likely the beads are pulling out what you want (methylated CpGs). That being said, I am not sure which correlation value is the more important one. The author of the program is quick to respond and was helpful for some of my questions.
Best of luck!



***
You are looking at genomic coverage essentially - whether the entire genome is covered by your reads or not. Again, the higher the value the better, but you'd want at least one. The way the math is done in the program, your reads are divided into two groups, A and B. The true correlation is the coverage using half your reads. The estimated correlation is half your reads but doubled... so essentially you would expect your true genome coverage to be somewhere in between. Most alignment programs will also give you a coverage value. The main thing is you want to make sure you are covering the reference ENOUGH, without overkill because that wastes money on sequencing. What value range you need really depends on your experiment. In our case we like to see at least 10x but less than 50x coverage, and if we are seeing too much coverage we will multiplex several samples into one lane of the sequencing plate.

I hope this helps! Sorry for the mix-up.

[NearyJL78]

Also, its the second number (.93-.96 in your case) that you might be looking at. That represents the reproducibility of the experiment with you current conditions... if you look at the graph you can get an idea how many reads you need to have to reproduce your experiment completely. If that number is too low (on a scale of 0-1) you might want to consider additional sequencing of the same material. In your data it looks to be sufficient in my opinion.