1
$\begingroup$

Relate to finding soft power, I ran "pickSoftThreshold" function and the result of that(sft) is as follow:

      "Power"   "SFT.R.sq"  "slope"   "truncated.R.sq"  "mean.k."   "median.k."     "max.k."
"1"     1         0.30       0.92         0.89           2056.1      1903.5          4558.9
"2"     2         0.50      -1.52         0.91            548.1       445.5          1973.9
"3"     3         0.78      -2.05         0.96            188.3       128.3          1015.9
"4"     4         0.81      -2.21         0.97            77.21       42.3            594.9
"5"     5         0.82      -2.21         0.98             36.1       16.2            373.2
"6"     6         0.80      -2.18         0.96             18.8       6.9             245.8
"7"     7         0.78      -2.07         0.94             10.6       3.1             167.8
"8"     8         0.93      -1.75         0.99             6.4        1.5             121.7
"9"     9         0.94      -1.74         0.99             4.2        0.79            103.0
"10"    10        0.95      -1.69         0.99            2.87        0.42            88.2
"11"    12        0.97      -1.60         0.98            1.52        0.13            66.4
"12"    13        0.96      -1.56         0.98            1.17        0.07            58.7
"13"    14        0.96      -1.52         0.98            0.92        0.04            52.3
"14"    15        0.96      -1.48         0.98            0.75        0.02            46.9
"15"    16        0.95      -1.45         0.98            0.62        0.01            42.2
"16"    17        0.95      -1.41         0.98            0.53        0.01            38.1
"17"    18        0.95      -1.36         0.99            0.45        0.01            34.6
"18"    19        0.94      -1.33         0.99            0.40        0.00            31.5

My network type is 'Signed Hybrid' and cor type is 'bicor' as follows:

sft = pickSoftThreshold(datExpr, powerVector = powers, networkType = "signed hybrid", corFnc="bicor" ,corOptions = list(maxPOutliers =0.1), verbose = 5)

based on the sft result that I presented in this post and also mean connectivity plot, I selected power = 7 to construct a co-expression network for 5000 most connected genes and 58 colerectal cancer patients. Now, I have 5 modules as below:

green     brown    grey   turquoise    yellow 
 550       345      3400       610        95 

As you see, 3400 genes are in the "grey" module. My question is that are these genes just in the "grey" module is reasonable? if it's not reasonable, how can I modify my constructing process? do the selected power as 7 is correct?

I really appreciated it if anybody share his/her comment with me.

$\endgroup$
3
  • $\begingroup$ How homogene is your data? All of your 58 samples are colorectal cancer patients? Isn't there any subtype of colon cancer? Why do you filter your data to the 5000 more connected? The tutorials clearly say you shouldn't filter the data (Besides how did you select that?). Is your data normalized or preprocessed in any way? $\endgroup$
    – llrs
    Dec 22, 2020 at 23:54
  • $\begingroup$ Thanks@llrs. my 58 samples are as one subtype of the colorectal cancer patients. most connected genes are acceptable filter because Professor Horvath used this filter before. my data was normalized through TPM method. could you please share your comment with me? $\endgroup$
    – Mohammad
    Dec 23, 2020 at 16:42
  • $\begingroup$ That someone used it in some analysis doesn't mean it is right with your analysis. You didn't share how did you filter those genes (or edit them on the question), I do not have further advice without further information $\endgroup$
    – llrs
    Jan 4, 2021 at 8:50

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.