Is there any possibility to use grn data from other software(grnboost2/GENIE3/CellOracle)?

[GilbertHan1011]

Hello,

After running my code I found that the transcription factor I wanted did not appear in the results. And since I have used some grn predict tools before, It may be better to use one tool uniformly. So I want to ask is there any possibility to use grn data from other software(grnboost2/GENIE3/CellOracle) for GRN result?

And I find that the GRN constructed is not cell-type-specific. Will this bring some bias?

Yours sincerely,
GIlbert

[edroaldo]

In principle, you can use the output of another GRN prediction tool. You can provide a data.frame containing your predicted regulatory interactions in which the first column contains the target gene and second column contains the regulator. See small example below (you can ignore the last two columns):

grn[1:5,]
target reg zscore corr
NOC2L CIART 4.207424 0.1505131
NOC2L DDIT3 4.678708 0.1403238
NOC2L DDX11 5.459027 0.1869369
NOC2L DICER1 4.848455 0.1674708
NOC2L DPF2 4.924815 0.1769888

Then, you can provide this table as input to the activeSignaling function (as described in the tutorial for spatial transcriptomics:
apathways <- activeSignaling(data=cellrouter@assays$ST@ndata, sampTab = cellrouter@assays$ST@sampTab, grn = grn, markers = markers, summary = summary, fc = 0.2)

Please, let me know if that works. Thanks!

[GilbertHan1011]

Thanks!

In software grnboost2, the result is like this

TF	target	importance
Hand2	Gm34030	314.14982364221123
Id1	Id3	272.82222256942646
Pou3f3	Pantr1	263.7302578553209
Irx5	Irx3	255.5759852168539
Irx3	Irx5	250.19502638751385
Zic4	Zic1	208.96672349106038
Nr2f2	Nr2f1	208.08884903096774
Id2	Id1	199.05869745900202
Kdm5d	Xist	194.6883118350258

In software cellOracle, the network is specific for clusters, which means different cluster have different network. the result is like this

,source,target,coef_mean,coef_abs,p,-logp
0,Zfp57,0610005C13Rik,8.765017e-05,8.765017e-05,0.48363494157656606,0.3154823298424891
1,Grhl2,0610005C13Rik,0.00020398656,0.00020398656,0.013902827081338852,1.8568968788117788
2,Zbtb7a,0610005C13Rik,-0.00037116688,0.00037116688,0.005543475024322045,2.2562179048095654
3,Foxf2,0610005C13Rik,-0.00078582985,0.00078582985,4.325799446150313e-07,6.363933619304989
4,Zfx,0610005C13Rik,0.0008235905,0.0008235905,4.678294839465598e-06,5.329912411189131
5,Nr5a1,0610005C13Rik,-9.254337e-06,9.254337e-06,0.9213348450624506,0.03558250342516251

So I think there are is some difficulty to use. I don't know whehter there are any ways to solve this.

Gilbert

[edroaldo]

Hello,
In principle you can use the grnboost2 network if you use the first two columns and rename the column names (potentially). I have not tested this and we are working on extensions based on your needs.

With regards the cellOracle network, I do not think there is a short-term solution. CellComm infers a global gene regulatory network exploring the heterogeneity of the data as a means to identify regulatory relationships. The, it idenfies subnetworks enriched in cell types depending on how the analysis is performed, as this is described in our tutorial.

Hope it helps. Thank you!