Regarding netmoss score and p-value and their interpretation

[ShailNair]

Hi,
My relative abundance table reads like this:

genus	control-1	control-2	case-1	case-2
Colwellia	0.51	0.35	43.3	56.4
Pseudoalteromonas	0.98	0.64	23.33	24.7
Cohaesibacter	0.07	0.05	0.07	0.1
xxxx	xx	xx	xx	xx

Netmoss results

taxon_names	NetMoss_Score	p.val	p.adj
Colwellia	1	0	0
Pseudoalteromonas	0.631	2.71E-147	6.53E-147
Cohaesibacter	1	0.607737938	0.61471341
....	...	...	..

Here the first two genera (Colwellia and Pseudoalteromonas) were highly abundant in CASE samples (relative abundance table). Once I run Netmoss I cannot see their significance (below image). Although Colwellia was enriched in CASE with a high netmoss score, the p-value is insignificant. On the other hand, Cohaesibacter which did not show much change in its abundance between the two samples (control vs case) shows significantly enriched (with the same netmoss score as that of Collwelia but zero log2FC).

here, which parameter (netmoss score, p-value (adj.) and Log2FC) is to be considered?

Also why Cohaesibacter in the above table have higher netmoss score and p-value than the other two genera (whose relative percentage increased highly in CASE samples)

Note: I run netmoss2 using default parameters

[xiaolw95]

NetMoss score is used to evaluate the difference in network structure between two groups, which may be not consistent with the difference in abundance. A higher NetMoss score refers to a greater difference in network structure but not in relative abundance. These two methods assess differential bacteria at different levels and to some extent, are complementary.

[ShailNair]

Thank you. What does "p.adj" mean here?. The top 9 genera (top to down) have similar NETMOSS scores. The first two genera were highly enriched in CASE. Then why does only Cohaesibacter is significant here?

[Damianyangyang]

Hi, My relative abundance table reads like this:

genus control-1 control-2 case-1 case-2
Colwellia 0.51 0.35 43.3 56.4
Pseudoalteromonas 0.98 0.64 23.33 24.7
Cohaesibacter 0.07 0.05 0.07 0.1
xxxx xx xx xx xx
Netmoss results

taxon_names NetMoss_Score p.val p.adj
Colwellia 1 0 0
Pseudoalteromonas 0.631 2.71E-147 6.53E-147
Cohaesibacter 1 0.607737938 0.61471341
.... ... ... ..

Here the first two genera (Colwellia and Pseudoalteromonas) were highly abundant in CASE samples (relative abundance table). Once I run Netmoss I cannot see their significance (below image). Although Colwellia was enriched in CASE with a high netmoss score, the p-value is insignificant. On the other hand, Cohaesibacter which did not show much change in its abundance between the two samples (control vs case) shows significantly enriched (with the same netmoss score as that of Collwelia but zero log2FC).

here, which parameter (netmoss score, p-value (adj.) and Log2FC) is to be considered?

Also why Cohaesibacter in the above table have higher netmoss score and p-value than the other two genera (whose relative percentage increased highly in CASE samples)

Note: I run netmoss2 using default parameters

Hello，Thank you. In the first table, is the data on diseases and health in a separate directory? When I run NetMoss it always have the error "Error: id variables not found in data: X".Have you encountered this problem?

[ShailNair]

@Damianyangyang Yes. the control group abundance file/s should be in the control control_dir while the case group abundance file/s should be in case_dir. Also, there should not be any duplicate values in the first column (taxa labels). see NetMoss2 test data for file setup.

Here is how I run it:

Load packages

library("NetMoss2")

#**Set seed**
sessionInfo()
set.seed(0168)

#Setup directories

setwd('~/Documents/Netmoss_analysis/')   ####set the working directory

#**read directory**
case_dir = paste0(getwd(),"/case")
control_dir = paste0(getwd(),"/control")
net_case_dir = paste0(getwd(),"/net_case_dir")
net_control_dir = paste0(getwd(),"/net_control_dir")

#contruct networks ####if files exist, skip

netBuild(case_dir = case_dir,
        control_dir = control_dir,
        method = "sparcc")

#calculate NetMoss score

nodes_result = NetMoss(case_dir = case_dir,    
        control_dir = control_dir,    
        net_case_dir = net_case_dir,   
        net_control_dir = net_control_dir) 

#Results

result = nodes_result[[1]]     ####NetMoss score result


write.table(result, "~/Documents//Netmoss_analysis/netmos.results.tsv", sep="\t", quote=F, col.names=NA)

#plot networks
netPlot(nodes_result)    ####image saved in working directory

[Damianyangyang]

@ShailNair Thank you very much. But I still have some questions. When I run it with some author's test data, NetMoss runs successfully. But when I use my own data, I still get this error.

And this is my own data. "net_case/control_dir" is generated with netBuild.
d_study1.txt
h_study1.txt
d_study_net1.txt
h_study_net1.txt

[ShailNair]

@Damianyangyang Did you check your results matrix table. When I ran your data, I got the following result:

Hence, it is clear why you get id variable not found in data :X.

Based on your data, it appears that you are using filtered data from a few species. It is strongly advised that you use the relative abundance table of all identified taxa. As you are using a relative abundance table, the data is mostly filtered for low variance and normalized for sequencing depth.

Or you can try with top 30-100 species. However, this is not advised since it will provide biased results.

[Damianyangyang]

@ShailNair Thank you very much for your guidance. I always have some errors when using the relative abundance data，but I can successfully run “netbuild” using the count of species. Is it necessary to run “netbuild” with the count of species？
And this is my data to run the nutbuild.
d_study1.txt
h_study1.txt

[ShailNair]

I always have some errors when using the relative abundance data

What error you are referring to? you can check the netmoss package document or the readme file for troubleshooting.

I could successfully run netmoss2 with your data:
d_study1.txt h_study1.txt

While plotting the interaction network, it showed
the threshold is too high to find the target taxon in the control networks!Error in if (as.character(edge.control2[mm, 1]) == top.tax2[kk] || as.character(edge.control2[mm, :

The default e.th is 0.4 . by adjusting it to 0.1 or lower you can visualize the plots.

Is it necessary to run “netbuild” with the count of species
Of course not if you already have the correlation data generated by other algorithms. In both cases, without the correlation data NetMoss cannot run

[Damianyangyang]

Thank you for your patience. I got these errors when I used the "netBuild" command to process the above two relative abundance data.
https://github.com/xiaolw95/NetMoss2/files/9131851/d_study1.txt
https://github.com/xiaolw95/NetMoss2/files/9131852/h_study1.txt
library("NetMoss2")
case_dir="/home/bioinfo/sunyy/work/NetMoss/netmoss_CRC/case_dir"
control_dir="/home/bioinfo/sunyy/work/NetMoss/netmoss_CRC/control_dir"
netBuild(case_dir=case_dir,control_dir=control_dir,method="sparcc")

[ShailNair]

@Damianyangyang This may come from conflicts in the R library or code or memory. If you ran the same command successfully with other data then it may probably be due to memory issues. I am not an expert in R to troubleshoot this, but I could successfully run your data on my 4 core-24GB RAM machine with NetMoss2 package

[Damianyangyang]

@ShailNair Thank you very much for your patient guidance.