Different results between LOLAweb interactive server and local R package

[NyxSLY]

I got different results from LOLAweb interactive server and LOLA installed from bioconductor with the same input files( region set, universe, and core database). What's the difference between web-based LOLA and LOLA R package?

[nsheff]

They should be the same. What kind of differences are you seeing?

[NyxSLY]

the most significant enrichment (a USCS feature table) in web-based LOLA is not significant at all in local R version.

web-based UCSC switchDbTss result (pValueLog is infinity):

userSet	dbSet	collection	pValueLog	oddsRatio	support	rnkPV	rnkLO	rnkSup	maxRnk	meanRnk	b	c	d	description	cellType	tissue	antibody	treatment	dataSource	filename	size	id	axis_label
allCT_regions.bed	3011	ucsc_features	25.14260683	470	1	76	10	76	29	112178	458	2748406	UCSC Switch DB TSSs					switchDbTss.bed	131780	UCSC Switch DB TSSs_3011	UCSC Switch DB TSSs

local R package UCSC switchDbTss result:
userSet dbSet collection pValueLog logOddsRatio support rnkPV rnkLO rnkSup maxRnk meanRnk b c d description cellType tissue antibody treatment dataSource filename qValue size
1 3011 ucsc_features 0.432783493920793 1.06557508871529 40 796 867 203 867 622 112504 995 2982142 UCSC Switch DB TSSs NA NA NA NA NA switchDbTss.bed 1 131780

I manually check the overlap between my regions and switchDbTss using findOverlaps function in R, and it seems the local version is right, there is not too many overlaps.

[nsheff]

Have you run this on the most recent version of lolaweb? can you try re-running it to see what happens? can you send me the link to the results on lolaweb?

[nsheff]

the d value is very different in your analyses which tells me the universe may be different. what reference genome assembly are you using? and what universe are you using?

[NyxSLY]

I just re-run LOLAweb with my own universe and you can see the results from http://lolaweb.databio.org/?key=W352OUYCHGVAK48 I also re-run LOLA in R and still see the difference. Input files, command and results are here https://drive.google.com/drive/folders/19pkIsQKCueKyqQh3OqqGa2vhhpMM29hs?usp=sharing The universe I am using is a 1kb tiles hg19 bed file made by myself. I have tried LOLAweb with premade universe tills1000.hg19.bed and my own 1kb tiles hg19 universe. Both of them give me almost the same results, but different with local R version. Thank you for your help!

…

On Wed, Apr 25, 2018 at 2:42 PM, Nathan Sheffield ***@***.***> wrote: the d value is very different in your analyses which tells me the universe may be different. what reference genome assembly are you using? and what universe are you using? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#27 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/APyCzu01MdhymFyrSeKpwNaUh94mM2KDks5tsNG3gaJpZM4Tj48n> .

[nsheff]

that is strange if you're using the same universe in both places. I will have to see if I can reproduce this. thanks for the files I'll take a look.

[NyxSLY]

Thanks!

…

On Wed, Apr 25, 2018 at 4:02 PM, Nathan Sheffield ***@***.***> wrote: that is strange if you're using the same universe in both places. I will have to see if I can reproduce this. thanks for the files I'll take a look. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#27 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/APyCzsWTJTpl0ZYe7_PssgjCpUjeEu5Nks5tsORVgaJpZM4Tj48n> .