Cellular deconvolution

Most cellular deconvolution algorithms consider an input data in a form of a matrix $X\_\{m\backslash times\; n\}$, which represents some molecular information (e.g. gene expression data or DNA methylation data) measured over a group of $n$ samples and $m$ marks (e.g. genes or CpG sites). The goal of the algorithm is to use these data and return an output matrix $W\_\{k\backslash times\; n\}$, representing the proportions of $k$ distinct cell types in each of the $n$ samples. Some methods limit the sum of each column of $W$ matrix less than or equal to one, so that the proportions of cell types sum up to the overall number of cells in the sample (less than one when there are some unknown cell types in the samples).{{cite journal | vauthors = Houseman EA, Kile ML, Christiani DC, Ince TA, Kelsey KT, Marsit CJ | title = Reference-free deconvolution of DNA methylation data and mediation by cell composition effects | journal = BMC Bioinformatics | volume = 17 | issue = 1 | pages = 259 | date = June 2016 | pmid = 27358049 | pmc = 4928286 | doi = 10.1186/s12859-016-1140-4 | doi-access = free }} Moreover, it is assumed that the values of $W$ matrix are non-negative as they pertain to proportions of cell types.

There are two broad categories of methods aiming at estimating the proportion of cell types in samples using some type of omics data (bulk gene expression or DNA methylation data). These approaches are labeled as reference-based (also called supervised) and reference-free (also called unsupervised) methods{{cite journal | vauthors = Teschendorff AE, Zheng SC | title = Cell-type deconvolution in epigenome-wide association studies: a review and recommendations | journal = Epigenomics | volume = 9 | issue = 5 | pages = 757–768 | date = May 2017 | pmid = 28517979 | doi = 10.2217/epi-2016-0153 | doi-access = free }}{{cite journal | vauthors = Sun X, Sun S, Yang S | title = An Efficient and Flexible Method for Deconvoluting Bulk RNA-Seq Data with Single-Cell RNA-Seq Data | journal = Cells | volume = 8 | issue = 10 | pages = 1161 | date = September 2019 | pmid = 31569701 | pmc = 6830085 | doi = 10.3390/cells8101161 | doi-access = free }}

Reference-based methods require an a priori defined reference matrix consisting of the expected value (also called profile or signature) of gene expression (or DNA methylation) for a group of genes (or CpG sites) known to have a differential expression (or methylation)

File:ReffreevsBased.png across the cell types. A reference matrix can be represented by a matrix $H\_\{m\backslash times\; k\}$, representing the expected value for $m$ markers (genes or CpG sites) for each of $k$ cell types known to be presented in the samples. These references can be derived by exploring external single-cell epigenomics or transcriptomics datasets generated for a group of samples similar (e.g. in terms of biological condition, sex and age) to the samples for which the deconvolution method will be applied. These methods use statistical approaches such as non-negative or constrained linear regression methods to dissect the contribution of each cell type to the aggregated bulk signals of genes or CpG sites.{{cite journal | vauthors = Titus AJ, Gallimore RM, Salas LA, Christensen BC | title = Cell-type deconvolution from DNA methylation: a review of recent applications | journal = Human Molecular Genetics | volume = 26 | issue = R2 | pages = R216–R224 | date = October 2017 | pmid = 28977446 | pmc = 5886462 | doi = 10.1093/hmg/ddx275 }} Constrained regression is the basis for many of reference-free cellular deconvolution methods existing in the literature, aiming at estimating the cell proportion values ($W\_\{k\backslash times\; n\}$) that maximizes the similarity between $HW^T$ and $X$. The performance of reference-based methods depends critically on the quality of the reference profiles.{{cite journal |last1=Sturm |first1=Gregor |last2=Finotello |first2=Francesca |last3=Petitprez |first3=Florent |last4=Zhang |first4=Jitao David |last5=Baumbach |first5=Jan |last6=Fridman |first6=Wolf H |last7=List |first7=Markus |last8=Aneichyk |first8=Tatsiana |title=Comprehensive evaluation of transcriptome-based cell-type quantification methods for immuno-oncology |journal=Bioinformatics |date=15 July 2019 |volume=35 |issue=14 |pages=i436–i445 |doi=10.1093/bioinformatics/btz363|pmid=31510660 |pmc=6612828 }}

There are a variety of approaches for isolating different cell types to measure their gene expression or DNA methylation levels to be used as references in the deconvolution algorithms. Earlier methods used cell sorting methods such as FACS (fluorescence-activated cell sorting) based on the flow cytometry technique, which separates the populations of cells belonging to different cell types based on their cell sizes, morphologies (shape), and surface protein expressions.{{cite journal | vauthors = Rosental B, Kozhekbaeva Z, Fernhoff N, Tsai JM, Traylor-Knowles N | title = Coral cell separation and isolation by fluorescence-activated cell sorting (FACS) | journal = BMC Cell Biology | volume = 18 | issue = 1 | pages = 30 | date = August 2017 | pmid = 28851289 | pmc = 5575905 | doi = 10.1186/s12860-017-0146-8 | doi-access = free }}{{cite journal | vauthors = Reinius LE, Acevedo N, Joerink M, Pershagen G, Dahlén SE, Greco D, Söderhäll C, Scheynius A, Kere J | display-authors = 6 | title = Differential DNA methylation in purified human blood cells: implications for cell lineage and studies on disease susceptibility | journal = PLOS ONE | volume = 7 | issue = 7 | pages = e41361 | date = 2012-07-25 | pmid = 22848472 | pmc = 3405143 | doi = 10.1371/journal.pone.0041361 | bibcode = 2012PLoSO...741361R | doi-access = free }}{{cite journal | vauthors = Koestler DC, Jones MJ, Usset J, Christensen BC, Butler RA, Kobor MS, Wiencke JK, Kelsey KT | display-authors = 6 | title = Improving cell mixture deconvolution by identifying optimal DNA methylation libraries (IDOL) | journal = BMC Bioinformatics | volume = 17 | issue = 1 | pages = 120 | date = March 2016 | pmid = 26956433 | pmc = 4782368 | doi = 10.1186/s12859-016-0943-7 | doi-access = free }} With the advance in single-cell technologies, newer approaches started to incorporate references for cell-types measured on a single-cell resolution obtained for a subset of subjects in the study or external subjects from a similar biological condition.{{cite journal | vauthors = Wang X, Park J, Susztak K, Zhang NR, Li M | title = Bulk tissue cell type deconvolution with multi-subject single-cell expression reference | journal = Nature Communications | volume = 10 | issue = 1 | pages = 380 | date = January 2019 | pmid = 30670690 | doi = 10.1038/s41467-018-08023-x | pmc = 6342984 | bibcode = 2019NatCo..10..380W }}{{cite journal | vauthors = Jew B, Alvarez M, Rahmani E, Miao Z, Ko A, Garske KM, Sul JH, Pietiläinen KH, Pajukanta P, Halperin E | display-authors = 6 | title = Accurate estimation of cell composition in bulk expression through robust integration of single-cell information | journal = Nature Communications | volume = 11 | issue = 1 | pages = 1971 | date = April 2020 | pmid = 32332754 | doi = 10.1038/s41467-020-15816-6 | pmc = 7181686 | bibcode = 2020NatCo..11.1971J }}

Reference-free methods do not need the reference profiles of cell-type specific genes (or CpGs), although they might still require the identity (name) of cell-type-specific genes (or CpGs).{{cite journal | vauthors = Tang D, Park S, Zhao H | title = NITUMID: Nonnegative matrix factorization-based Immune-TUmor MIcroenvironment Deconvolution | journal = Bioinformatics | volume = 36 | issue = 5 | pages = 1344–1350 | date = March 2020 | pmid = 31593244 | doi = 10.1093/bioinformatics/btz748 | pmc = 8215918 }} These methods might be considered as a modification of reference-based methods where both $H$ and $W$ are unknown, and the goal is to jointly estimate both matrices so that the similarity between $HW^T$ and $X$ is maximized. Many of the reference-free methods are based on mathematical framework of non-negative matrix factorization,{{cite journal | vauthors = Repsilber D, Kern S, Telaar A, Walzl G, Black GF, Selbig J, Parida SK, Kaufmann SH, Jacobsen M | display-authors = 6 | title = Biomarker discovery in heterogeneous tissue samples -taking the in-silico deconfounding approach | journal = BMC Bioinformatics | volume = 11 | issue = 1 | pages = 27 | date = January 2010 | pmid = 20070912 | doi = 10.1186/1471-2105-11-27 | pmc = 3098067 | doi-access = free }}{{cite journal | vauthors = Moffitt RA, Marayati R, Flate EL, Volmar KE, Loeza SG, Hoadley KA, Rashid NU, Williams LA, Eaton SC, Chung AH, Smyla JK, Anderson JM, Kim HJ, Bentrem DJ, Talamonti MS, Iacobuzio-Donahue CA, Hollingsworth MA, Yeh JJ | display-authors = 6 | title = Virtual microdissection identifies distinct tumor- and stroma-specific subtypes of pancreatic ductal adenocarcinoma | journal = Nature Genetics | volume = 47 | issue = 10 | pages = 1168–78 | date = October 2015 | pmid = 26343385 | doi = 10.1038/ng.3398 | pmc = 4912058 }} which imposes a non-negativity constraint on the elements of $H$ and $W$. Additional constraints such as the assumption of orthogonality between the columns of $H$ might be incorporated to improve the interpretability of results and prevent overfitting.

The advance of single-cell technologies enables the profiling of each individual cell in a sample, which help elucidate the issue of cellular heterogeneity by measuring the proportions of different cells in samples. Even though the quality of single cell profiling technologies has been on the rise in recent years, these technologies are still costly, limiting their applications in large populations of samples.{{cite journal | vauthors = Wang X, Park J, Susztak K, Zhang NR, Li M | title = Bulk tissue cell type deconvolution with multi-subject single-cell expression reference | journal = Nature Communications | volume = 10 | issue = 1 | pages = 380 | date = January 2019 | pmid = 30670690 | pmc = 6342984 | doi = 10.1038/s41467-018-08023-x | bibcode = 2019NatCo..10..380W }} Single cell technologies such as single cell transcriptomic methods also tend to have higher error rates due to factors such as high dropout events.{{cite journal | vauthors = Ran D, Zhang S, Lytal N, An L | title = scDoc: correcting drop-out events in single-cell RNA-seq data | journal = Bioinformatics | volume = 36 | issue = 15 | pages = 4233–4239 | date = August 2020 | pmid = 32365169 | doi = 10.1093/bioinformatics/btaa283 }}{{cite journal | vauthors = Yamawaki TM, Lu DR, Ellwanger DC, Bhatt D, Manzanillo P, Arias V, Zhou H, Yoon OK, Homann O, Wang S, Li CM | display-authors = 6 | title = Systematic comparison of high-throughput single-cell RNA-seq methods for immune cell profiling | journal = BMC Genomics | volume = 22 | issue = 1 | pages = 66 | date = January 2021 | pmid = 33472597 | pmc = 7818754 | doi = 10.1186/s12864-020-07358-4 | doi-access = free }} Cellular deconvolution methods provide a robust and cost-effective in silico alternatives for understanding the samples on a cell-type level resolution, by relying on single cell information of only a small subset of cells in the sample, the reference profiles generated by external sources, or even no reference profile at all.{{cite journal | vauthors = Wang J, Roeder K, Devlin B | title = Bayesian estimation of cell-type-specific gene expression per bulk sample with prior derived from single-cell data. | journal = bioRxiv | date = January 2020 | doi = 10.1101/2020.08.05.238949 | s2cid = 221096767 }}

There are large amounts of old bulk data, such as microarray, from studies concerning various diseases and biological conditions. These datasets could be considered important resources in studying of rare disease, long follow-up studies or samples and tissues that are difficult to extract. In addition, this can also improve the statistical power by combining similar datasets.{{Cite journal |last1=Yau |first1=Tung On |last2=Vadakekolathu |first2=Jayakumar |last3=Foulds |first3=Gemma Ann |last4=Du |first4=Guodong |last5=Dickins |first5=Benjamin |last6=Polytarchou |first6=Christos |last7=Rutella |first7=Sergio |date=March 2022 |title=Hyperactive neutrophil chemotaxis contributes to anti-tumor necrosis factor-α treatment resistance in inflammatory bowel disease |journal=Journal of Gastroenterology and Hepatology |language=en |volume=37 |issue=3 |pages=531–541 |doi=10.1111/jgh.15764 |issn=0815-9319 |pmc=9303672 |pmid=34931384}} Since the biological samples for many of these studies are not available or accessible anymore, reprofiling the data using single cell technologies might not be within the realm of possibilities for many studies. Invention of more advanced cellular deconvolution methods gives the opportunity to researchers to come back to old omics studies, reanalyze their datasets, and scrutinize their findings.

Reference-based approaches rely on the availability of accurate references to estimate cell proportions. The discrepancy between the biology of the samples underlying the references and the samples for which the cell proportions are being estimated could introduce bias in estimated cell proportions.{{cite journal | vauthors = Gervin K, Salas LA, Bakulski KM, van Zelm MC, Koestler DC, Wiencke JK, Duijts L, Moll HA, Kelsey KT, Kobor MS, Lyle R, Christensen BC, Felix JF, Jones MJ | display-authors = 6 | title = Systematic evaluation and validation of reference and library selection methods for deconvolution of cord blood DNA methylation data | journal = Clinical Epigenetics | volume = 11 | issue = 1 | pages = 125 | date = August 2019 | pmid = 31455416 | pmc = 6712867 | doi = 10.1186/s13148-019-0717-y | doi-access = free }} Studies have shown that using references obtained from samples with different phenotypes such as age, gender, and disease status than the population of interest reduces the performance of reference-based methods to levels lower than their reference-free counterparts.

Reference-based approaches assume the existence of prior knowledge on the types of cells existing in a sample. Therefore, these methods may fail to perform accurately when the data includes rare or otherwise unknown cell types with no references incorporated in the algorithm. For example, cancer tumors consist of heterogeneous mixtures of various healthy cells of different types such as immune cells and cells related to affected tissues in addition to tumor cells. Although it might be possible to provide references for the immune cells, we do not usually have access to references or signatures for cancer cells due to the unique patterns of mutations and distributions of molecular information in each individual. These situations have been addressed in some studies under the label of deconvolution methods with partial reference availability.{{cite journal | vauthors = Qin Y, Zhang W, Sun X, Nan S, Wei N, Wu HJ, Zheng X | title = Deconvolution of heterogeneous tumor samples using partial reference signals | journal = PLOS Computational Biology | volume = 16 | issue = 11 | pages = e1008452 | date = November 2020 | pmid = 33253170 | pmc = 7728196 | doi = 10.1371/journal.pcbi.1008452 | bibcode = 2020PLSCB..16E8452Q | doi-access = free }}

File:ConfoundingAlzS.png

Studies have shown that the proportions of different cell types might show correlations with various phenotypes such as different diseases. For example, the proportions of Parathyroid oxyphil cells in the samples collected from the parathyroid gland for groups of patients show a significant correlation with the presence of clinical characteristics of chronic kidney disease (CKD).{{cite journal | vauthors = Ding Y, Zou Q, Jin Y, Zhou J, Wang H | title = Relationship between parathyroid oxyphil cell proportion and clinical characteristics of patients with chronic kidney disease | journal = International Urology and Nephrology | volume = 52 | issue = 1 | pages = 155–159 | date = January 2020 | pmid = 31686279 | doi = 10.1007/s11255-019-02330-y | s2cid = 207895174 }} Another study applying the cellular deconvolution algorithms to gene expression data of Alzheimer's patients find that patients with lower proportions of neuronal cells in the samples collected from their cerebral cortex are more likely to show the clinical characteristics of dementia.{{cite journal | vauthors = Andrade-Moraes CH, Oliveira-Pinto AV, Castro-Fonseca E, da Silva CG, Guimarães DM, Szczupak D, Parente-Bruno DR, Carvalho LR, Polichiso L, Gomes BV, Oliveira LM, Rodriguez RD, Leite RE, Ferretti-Rebustini RE, Jacob-Filho W, Pasqualucci CA, Grinberg LT, Lent R | display-authors = 6 | title = Cell number changes in Alzheimer's disease relate to dementia, not to plaques and tangles | journal = Brain | volume = 136 | issue = Pt 12 | pages = 3738–52 | date = December 2013 | pmid = 24136825 | doi = 10.1093/brain/awt273 | pmc = 3859218 }} Cellular deconvolution algorithms could enable researchers to investigate the interactions between cell proportions and various diseases or biological phenotypes.

Epigenome-wide association study (EWAS) and transcriptome-wide association studies (TWAS) aim at finding the molecular markers such as genes or methylation CpG sites that show significant correlations between their expression or methylation levels and a biological phenotype of interest such as a disease. Since the proportions of cell types in samples vary and might show significant correlations with the disease or phenotype of interest, these correlations may confound the functional relationships between genes or CpG sites and the disease or phenotypes under study.{{cite journal | vauthors = Glastonbury CA, Couto Alves A, El-Sayed Moustafa JS, Small KS | title = Cell-Type Heterogeneity in Adipose Tissue Is Associated with Complex Traits and Reveals Disease-Relevant Cell-Specific eQTLs | journal = American Journal of Human Genetics | volume = 104 | issue = 6 | pages = 1013–1024 | date = June 2019 | pmid = 31130283 | doi = 10.1016/j.ajhg.2019.03.025 | pmc = 6556877 }} For example, studies aimed at finding genes involved in Alzheimer's disease may end up selecting genes that are exclusively expressed in neurons and therefore have lower expression levels in Alzheimer's patients due to compositional changes of cell types during neurodegeneration.{{cite journal | vauthors = Mathys H, Davila-Velderrain J, Peng Z, Gao F, Mohammadi S, Young JZ, Menon M, He L, Abdurrob F, Jiang X, Martorell AJ, Ransohoff RM, Hafler BP, Bennett DA, Kellis M, Tsai LH | display-authors = 6 | title = Single-cell transcriptomic analysis of Alzheimer's disease | journal = Nature | volume = 570 | issue = 7761 | pages = 332–337 | date = June 2019 | pmid = 31042697 | pmc = 6865822 | doi = 10.1038/s41586-019-1195-2 | bibcode = 2019Natur.570..332M }} Such genes are not actionable targets for the treatment of Alzheimer's since they are not causally involved in the biological mechanism underlying Alzheimer's disease, but are only brought up by the confounding effects of cell types.

{{Reflist}}

Category:Developmental biology