bidirectional motifs and positions

[KevCaz]

Would be good to have another function that handle bidirectional network for 3 nodes motifs.
It'd mean to handle 64 cases, quite boring to write but there might be a way of writing it smartly using boolean algebra!

edit ✏️
What remains is:

• enhance the doc of the function motif_census(), especially to list the position for the bidirecitonal case;
• add a count of motif per species;
• add a total count of positions.
• add test

[KevCaz]

I'll follow the motifs order as in fig. 1 in Milo 2002, for each motif I'll do every position so that we have a clear mapping between position and motif and so it'll be easy to get a count of motif based on the count of positions.

[KevCaz]

Depending on the performance it'd may be make sense to use the function even for unidirectional network.

[david-beauchesne]

Why not do it for 4-species motifs as well?

[KevCaz]

too many positions (256).... unless there is a smart way to proceed I won't do that 😆 !

[david-beauchesne]

👍 haha fair enough!

[KevCaz]

See d0556c4

[KevCaz]

Performance-wise there is still a small advantage using the unidirectional implementation

R> nsp <- 200 
   mat <- matrix(runif(nsp^2)>.5, nsp, nsp) 
    
   microbenchmark( 
       res2 <- motif_census(mat, unidirectional = TRUE), 
       res <- motif_census(mat),  
       times = 50 
   )                                                             

Unit: milliseconds
                                             expr      min       lq     mean   median       uq      max neval
 res2 <- motif_census(mat, unidirectional = TRUE)  99.3449 102.5198 103.6208 103.5854 104.9859 108.9092    50
                         res <- motif_census(mat) 123.1422 128.6429 132.1059 130.3861 133.0326 174.4492    50