LuxGLM: a probabilistic covariate model for quantification of DNA methylation modifications with complex experimental designs

October 21, 2024 · View on GitHub

Overview

LuxGLM is a method for quantifying oxi-mC species with arbitrary covariate structures from bisulphite based sequencing data. LuxGLM's probabilistic modeling framework combines a previously proposed hierarchical generative model of Lux for oxi-mC-seq data and a general linear model component to account for confounding effects.

Features

Model-based integration and analysis of BS-seq/oxBS-seq/TAB-seq/fCAB-seq/CAB-seq/redBS-seq/MAB-seq/etc. data from whole genome, reduced representation or targeted experiments
Accounts for confounding effects through general linear model component
Considers nonideal experimental parameters through modeling, including e.g. bisulphite conversion and oxidation efficiencies, various chemical labeling and protection steps etc.
Model-based integration of biological replicates
Detects differential methylation using Bayes factors (DMRs)
Full Bayesian inference using NumPyro

Quick introduction

An usual LuxGLM pipeline has the following steps

Alignment of BS-seq and oxBS-seq data (e.g., Bismark or BSmooth)
Extraction of converted and unconverted counts (e.g., Bismark or BSmooth)
Integrative methylation analysis
Analysis of obtained methylation estimates, e.g., using Bayes factors

This documentation focuses on the third and fourth points.

Installation

PyPI

$ pip install luxglm

GitHub

Install the version from the main branch as follows

$ pip install git+https://github.com/tare/LuxGLM.git

Usage

Metadata

Count data and covariates are defined in the metadata file.

name	basal/tgf-beta	timepoint	count_file	control_count_file	control_definition_file
TGFb_1_24h	1	24	wildtype/TGFb_1_24h.tsv	control/TGFb_1_24h_control.tsv	control_definitions.tsv
TGFb_1_38h	1	38	wildtype/TGFb_1_38h.tsv	control/TGFb_1_38h_control.tsv	control_definitions.tsv
TGFb_1_48h	1	48	wildtype/TGFb_1_48h.tsv	control/TGFb_1_48h_control.tsv	control_definitions.tsv

The following columns are mandatory: name, count_file, control_count_file, and control_definition. Additionally, there has to be at least one covariate. In the above example, we have four covariates: basal/tgf-beta, vitc, ra, and timepoint.

Control cytosines

The control cytosine data are supplied in the control count files. Each experiment will have its own file. The files contain location information (chromosome and position) and control type information (control_cytosine) for the control cytosines. Additionally, we have the number of Cs and and total number of read-outs from BS-seq and oxBS-seq experiments (bs_c, bs_total, oxbs_c, and oxbs_total).

chromosome	position	control_type	bs_c	bs_total	oxbs_c	oxbs_total
Lambda_ctrl	22924	C	2	343	1	562
Lambda_ctrl	22928	C	2	341	1	561
Lambda_ctrl	47359	5mC	3770	3857	4767	4877
Lambda_ctrl	47367	5mC	3895	3962	4855	4979
Lambda_ctrl	23789	5hmC	3792	3964	79	865
Lambda_ctrl	23794	5hmC	3901	4115	62	934

The prior knowledge on the control cytosines is supplied in the control definition file. Note that control_type is used to link the control count data and control definitions.

control_type	C_pseudocount	5mC_pseudocount	5hmC_pseudocount
C	998	1	1
5mC	1	998	1
5hmC	6	2	72

Noncontrol cytosines

The non-control cytosine data are supplied in the count files. Each experiment will have its own file. The files contain location information (chromosome and position) for the non-control cytosines. Additionally, we have the number of Cs and and total number of read-outs from BS-seq and oxBS-seq experiments (bs_c, bs_total, oxbs_c, and oxbs_total).

chromosome	position	bs_c	bs_total	oxbs_c	oxbs_total
chrX	7159069	1083	1563	850	2736
chrX	7159186	1341	1534	2119	2719
chrX	7159222	4949	5575	3886	4639
chrX	7159235	4831	5588	4354	4641

LuxGLM analysis

The following lines are sufficient to run LuxGLM

import numpyro
from jax import random
from luxglm.inference import run_nuts
from luxglm.utils import get_input_data

numpyro.enable_x64()

# read input data
lux_input_data = get_input_data("metadata.tsv")

key = random.PRNGKey(0)
key, key_ = random.split(key)

# run LuxGLM
lux_result = run_nuts(
    key,
    lux_input_data,
    ["basal/tgf-beta"],
    num_warmup=1_000,
    num_samples=1_000,
    num_chains=4,
)

# ensure convergence
lux_result.inference_metrics["summary"].query("r_hat > 1.05")

To get the posterior samples of methylation levels of control and non-control cytosines one can call lux_result.methylation_controls() and lux_result.methylation(), respectively.

The posterior samples of experimental parameters can be obtained by calling lux_result.experimental_parameters().

To study the effects of the covariates, one can get the posterior samples of coefficients of covariates using lux_result.coefficients().

Examples

Please see the examples directory for the tutorial notebooks.

References

[1] T. Äijö, X. Yue, A. Rao and H. Lähdesmäki, “LuxGLM: a probabilistic covariate model for quantification of DNA methylation modifications with complex experimental designs.,” Bioinformatics, 32.17:i511-i519, Sep 2016.

[2] T. Äijö, Y. Huang, H. Mannerström, L. Chavez, A. Tsagaratou, A. Rao and H. Lähdesmäki, “A probabilistic generative model for quantification of DNA modifications enables analysis of demethylation pathways.,” Genome Biol, 17.1:1, Mar 2016.

[3] F. Krueger and S. R. Andrews, “Bismark: a flexible aligner and methylation caller for Bisulfite-Seq applications.,” Bioinformatics, 27.11:1571-1572, Jun 2011.

[4] K. D. Hansen, B. Langmead and R. A. Irizarry, “BSmooth: from whole genome bisulfite sequencing reads to differentially methylated regions.,” Genome Biol, 13.10:1, Oct 2012.

[5] D. Phan, N. Pradhan and M. Jankowiak, “Composable Effects for Flexible and Accelerated Probabilistic Programming in NumPyro.,” arXiv preprint 1912.11554, Dec 2019.