PPubMed20 Oct 2021Cited 34×

Bayesian log-normal deconvolution for enhanced in silico microdissection of bulk gene expression data

Bárbara Andrade Barbosa, Saskia D. van Asten, Ji Won Oh, Arantza Farina-Sarasqueta, Joanne Verheij, Frederike Dijk, Hanneke W. M. van Laarhoven, Bauke Ylstra, Juan J. Garcia Vallejo, Mark A. van de Wiel, Yongsoo Kim

Abstract

'Arantza Farina-Sarasqueta' 'Joanne Verheij' 'Frederike Dijk' 'Hanneke W. M. van Laarhoven' 'Bauke Ylstra' 'Juan J. Garcia Vallejo' 'Mark A. van de Wiel' 'Yongsoo Kim'] Deconvolution of bulk gene expression profiles into the cellular components is pivotal to portraying tissue’s complex cellular make-up, such as the tumor microenvironment. However, the inherently variable nature of gene expression requires a comprehensive statistical model and reliable prior knowledge of individual cell types that can be obtained from single-cell RNA sequencing. We introduce BLADE (Bayesian Log-normAl Deconvolution), a unified Bayesian framework to estimate both cellular composition and gene expression profiles for each cell type. Unlike previous comprehensive statistical approaches, BLADE can handle > 20 types of cells due to the efficient variational inference. Throughout an intensive evaluation with > 700 simulated and real datasets, BLADE demonstrated enhanced robustness against gene expression variability and better completeness than conventional methods, in particular, to reconstruct gene expression profiles of each cell type. In summary, BLADE is a powerful tool to unravel heterogeneous cellular activity in complex biological systems from standard bulk gene expression data.

A figure from Bayesian log-normal deconvolution for enhanced in silico microdissection of bulk gene expression data — fig. from the paper

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Answer

BLADE applies a log‑normal hierarchical Bayesian model to bulk RNA‑seq, delivering more accurate cell‑type fractions and purified expression profiles with uncertainty estimates.

What’s new

i.Introduced BLADE, a hierarchical Bayesian model that jointly deconvolutes bulk RNA‑seq and estimates cell‑type‑specific expression profiles
ii.Implemented a log‑normal convolution to accurately capture gene‑expression variability unlike traditional normal‑based methods
iii.Integrated prior knowledge from scRNA‑seq (means and variances) to constrain and improve inference
iv.Provided posterior uncertainty for both cell‑type fractions and purified expression, enabling downstream confidence assessment
v.Enabled efficient variational inference allowing analysis of many cell types simultaneously

How they did it

BLADE builds a hierarchical Bayesian graphical model where bulk gene counts are modeled as a log‑normal mixture of cell‑type expression profiles; prior means and variances are derived from scRNA‑seq (e.g., CITE‑seq) and variational inference is used to estimate posterior distributions of cell fractions and per‑cell‑type expression.

What they found

In simulated data with up to ten cell types, BLADE achieved Pearson r > 0.9 for fraction estimates, outperforming NNLS (r ≈ 0.6) and CIBERSORTx (r ≈ 0.8) (Fig 4a)
Across PBMC and tumor simulations, BLADE consistently gave higher correlation for purified gene expression (up to 0.95) than competing methods (Fig 5b, Fig 7b)
When prior knowledge was incomplete, BLADE remained robust, retaining superior performance over CIBERSORTx and MuSiC (Fig 6b)
Applied to PDAC bulk data, BLADE’s fraction estimates correlated best with flow‑cytometry (highest Pearson coefficients) and recovered more unique DEGs per cell type than alternatives

Caveats

✕Requires a reference scRNA‑seq or CITE‑seq dataset; performance degrades if the reference is missing or inaccurate
✕Assumes gene‑wise log‑normality, which may not hold for all genes or extreme expression regimes
✕Variational inference, while fast, provides approximate posteriors and may miss multimodal solutions
✕Validation focused on simulated mixtures and specific cancer datasets; broader tissue applicability remains to be demonstrated

in plain words

Imagine a soup made of many different ingredients, but you only have a picture of the whole bowl. BLADE uses clever math that assumes each ingredient’s flavor follows a log‑normal pattern, and it also uses clues from single‑cell data to figure out how much of each ingredient is in the soup and what each ingredient tastes like. This lets scientists study each cell type without actually separating them first.

why it matters

Accurate in‑silico microdissection lets researchers extract cell‑type‑specific signals from routine bulk RNA‑seq, enabling detailed tumor‑microenvironment profiling, biomarker discovery, and cost‑effective studies without expensive single‑cell experiments.

Follow the thread

deconvolution log‑normal hierarchical Bayesian BLADE bulk RNA‑seq cell‑type fraction

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

Similar papers in all four archives

The same idea searched in arXiv, PubMed, bioRxiv and chemRxiv at once, so a method can lead you to where it is being applied.

AarXiv4 found

Searching the other archives…