FLAMINGOrLite

Lite version of FLAMINGO, high-resolution 3D chromosome structures reconstruction based on Hi-C.

Now only support .hic file. .mcool support is coming soon.

Dependencies

The implementation of the algorithm is based on R. It depends on 3 basic R packages: strawr, parallel, and Matrix.

Installation of the FLAMINGOrLite package

install.packages("devtools")
library(devtools)
install_github('JiaxinYangJX/FLAMINGOrLite',ref='HEAD')

Differences between FLAMINGOrLite and FLAMINGOr

1. FLAMINGOrLite doesn't support iFLAMINGO
2. FLAMINGOrLite is faster and more memory-efficient
3. FLAMINGOrLite is more user-friendly and fixed bugs in FLAMINGOr

Reconstruct the 3D genome structure with FLAMINGOrLite

Example:

Using 20 threads to reconstruct 5kb-resolution (frag_res) 3D structure of chr21.

The resolution of skeleton (domain_res) is set to be 1Mb, which means one domain contains 200 fragments (frag_res / domain_res).

library(FLAMINGOrLite)
res = flamingo_main(hic_data='4DNFI1UEG1HD.hic',
                    file_format='hic',
                    domain_res=1e6,
                    frag_res=5e3,
                    chr_name='chr21',
                    normalization='KR',
                    nThread=20)

Key functions

If the genome comprises over 200 fragments, we strongly recommend users employ flamingo_main rather than flamingo_basic for hierarchical structure reconstruction. Users should carefully choose appropriate values for domain_res and frag_res to strike a balance between the number of domains in the skeleton (genome size / domain_res) and the number of fragments within each domain (domain_res / frag_res).

---

flamingo_main

Main function of FLAMINGO, hieractically reconstruct 3D genome structure.

This function can be used to efficiently predict the structure of high-resolution whole chromosome structures.

It takes three steps:

1. Reconstruct low-resolution (domain_res) genome skeleton based on flamingo_basic
2. Reconstruct high-resolution (frag_res) intra-domain structures based on flamingo_basic in parallel
3. Assembly the high-resolution domains into the genome skeleton

Output:

col	abbrev	Description
1	chr	chromosome ID
2	start	genomic location of the start point
3	end	genomic location of the end point
4	x	x coordinate
5	y	y coordinate
6	z	z coordinate

Type ?flamingo_main for detailed explanations of each argument.

---

flamingo_basic

Core function of FLAMINGO, 3D genome structure reconstruction using low-rank matrix completion.

It only takes interaction frequecy matrix as the input.

This function can be used to predict the structure within 200 fragments.

res = flamingo_basic(input_if)

Output:

A flamingo_prediction object containing the fragment id and 3D coordinates

Type ?flamingo_basic for detailed explanations of each argument.

visualize the 3D genome structure using ParaView

ParaView is an open-source, multi-platform data analysis and visualization application. To visualize the 3D genome structure using FLAMINGO, the user needs to convert the 3D coordinates into a .vtk file and use the ParaView software to visualize the structure. In the FLAMINGOrLite package, a write.vtk function is provided.

write.vtk(points,lookup_table,name,opt_path)

Arguments:

points: 3D coordinates predicted by FLAMINGO in the x,y,z format. 3 by N matrix.

lookup_table: The annotation of each point, could be labels or scores, i.e. the compartment PC scores. N-dimensional vector

name: output file name annotated within the file. String.

opt_path: output file path including the file name. String.

Output:

A .vtk file stored at opt_path for ParaView visualization.

Type ?write.vtk for detailed explanations of each argument.