Cell and Tissue Enrichment Methods and Resources
1 Bulik-Sullivan et al. 2015 β LD Score Regression (LDSC)
Key Findings::
- Developed LD Score regression to distinguish true polygenic signal from confounding bias (cryptic relatedness, population stratification)
- LDSC intercept quantifies inflation from each source
- More powerful and accurate than genomic control for GWAS correction
- Foundation for subsequent stratified LDSC (Finucane 2018) used in tissue/cell-type enrichment
- Validated on 22 GWAS datasets including schizophrenia, height, BMI, diabetes
- Highly cited (6,174 citations); adopted as standard GWAS QC pipeline
Conclusion:: LDSC is a foundational method enabling downstream tissue/cell-type enrichment by reliably separating polygenicity from confounding. Relevance:: LDSC is the mathematical core upon which stratified LDSC (S-LDSC) for cell-type enrichment is built. Used by DecodeME, Maccallini 2026, mecfsscience.org for ME/CFS enrichment analyses. Without LDSC, tissue enrichment from GWAS summary statistics would be unreliable. Quality:: Very High (Nature Genetics, 6,174 citations, peer-reviewed, standard tool) Certainty:: 0.92 Limitations:: Assumes infinitesimal polygenic architecture; sensitive to LD reference panel choice; does not itself perform enrichment β only partitioning.
2 de Leeuw et al. 2015 β MAGMA Gene-Set Analysis
Leeuw et al. (2015)
Key Findings::
- Developed MAGMA (Multi-marker Analysis of GenoMic Annotation) for gene and gene-set analysis of GWAS data
- Gene analysis via multiple regression model β significantly better power than single-SNP methods
- Gene-set analysis built as separate layer with regression structure enabling continuous gene properties
- Outperforms PLINK, ALIGATOR, INRICH, MAGENTA in simulation and Crohn's disease benchmarks
- Competitive gene-set analysis: is a gene set more associated than the rest of the genome?
- Corrects for LD, gene size, gene density; dramatically faster than permutation-based methods
- >2,000 citations; adopted as default gene-set method in FUMA platform
Conclusion:: MAGMA is the standard tool for gene-set and cell-type enrichment from GWAS β the engine behind most ME/CFS enrichment results. Relevance:: MAGMA is used in DecodeME tissue analysis, Maccallini 2026 cell-type enrichment, the Duncan 2025 pipeline, and the mecfsscience.org analyses. Understanding its methodology is essential to evaluating the ME/CFS enrichment findings. Quality:: Very High (PLoS Comp Biol, >2,000 citations, peer-reviewed, platform standard) Certainty:: 0.90 Limitations:: Gene boundary definition (window size) affects results; reference LD panel can bias results; no native trans-ethnic support; competitive gene-set analysis sensitive to background gene set selection.
3 Finucane et al. 2018 β Stratified LDSC for Disease-Relevant Tissue/Cell-Type Enrichment
Key Findings::
- Introduced stratified LD score regression (S-LDSC) to partition SNP heritability by functional annotation
- Created 10 cell-type-specific gene sets from GTEx v6, ImmGen, Franke lab RNA data
- Validated on 48 diseases/traits β correctly identified known disease biology:
- Immune cells for rheumatoid arthritis, Crohn's disease, asthma
- CNS/neuronal for schizophrenia, epilepsy
- Pancreatic cells for diabetes
- Published annotation files as Google Cloud public resource
- Pipeline used directly by mecfsscience.org for their ME/CFS analysis (analysis 1)
Conclusion:: S-LDSC is a validated method that correctly identifies disease-relevant cell types from GWAS data. Applying it to ME/CFS yields a strong CNS/neuronal signal. Relevance:: The mecfsscience.org blog applied the exact Finucane 2018 pipeline to DecodeME results and found strong CNS enrichment for ME/CFS. Validates approach: method correctly found immune cells for RA/Crohnβs and neural cells for schizophrenia β same pipeline applied to ME/CFS identifies neurons. Quality:: Very High (Nature Genetics, peer-reviewed, foundation of the entire GWAS cell-type enrichment field) Certainty:: 0.90 Limitations:: Only 10 broad categories β cannot resolve closely related cell subtypes within tissue; method performance degrades for fine annotations; assumes SNPs near specifically expressed genes tag relevant regulatory variants; finer-grained analysis requires cell-type-specific RNA datasets (HBA, Dropviz) and MAGMA-based methods.
4 Skene et al. 2018 β Brain Cell Type Enrichment for Schizophrenia (Linnarsson Mouse Atlas)
Key Findings::
- Applied MAGMA cell-type enrichment to schizophrenia GWAS (PGC2: 40,675 cases, 64,643 controls) using Linnarsson mouse brain snRNA-seq atlas (9,970 cells)
- Schizophrenia genetic signal significantly enriched in: CA1 pyramidal neurons, subiculum neurons, medium spiny neurons (D1 and D2 subtypes), somatostatin-positive interneurons
- Validation across 7 independent cell-type datasets showed consistent neuronal enrichment
- Published analysis pipeline via R package `MAGMA_celltyping` (still widely used)
- Level 1 taxonomy (~70 broad cell types) used by mecfsscience.org analysis 7
Conclusion:: Demonstrates that MAGMA + single-cell atlases can identify specific brain cell types underlying psychiatric disease risk. MSN-related findings for schizophrenia prefigure the MSN enrichment now observed in ME/CFS. Relevance:: Pipeline and methodology directly transferred to ME/CFS by mecfsscience.org. The MSN finding for schizophrenia establishes precedent: MSN enrichment is a biologically meaningful signal, not an artifact of the method. However, MSN enrichment is shared across multiple brain traits β not specific to any single condition. Quality:: Very High (Nature Genetics, peer-reviewed, highly cited, published package) Certainty:: 0.88 Limitations:: Mouse-to-human gene mapping introduces noise; limited to ~70 broad cell-type categories; cross-species cell-type homology assumptions; Level 2/3 fine cell types require higher-resolution atlases.
5 Saunders et al. 2018 β Dropviz Mouse Brain Atlas (Reference Resource)
Key Findings::
- Drop-seq snRNA-seq atlas: 690,000 single-nucleus transcriptomes from 9 regions of adult mouse brain
- Identified 565 transcriptionally distinct cell populations
- Striatum contains functionally diverse MSN subtypes including eMSN-like cells
- Gene expression profiles accessible via FUMA download page for cross-species GWAS enrichment
- Naming caveat: GP (globus pallidus) neuron clusters actually represent striatal cell types (documented in paper)
- Atlas validated by ISH cross-reference with Allen Brain Atlas
Conclusion:: Dropviz is a foundational reference for cross-species cell-type enrichment β the primary mouse dataset used by Maccallini 2026 to identify striatal neuron enrichment in ME/CFS. Relevance:: In Maccallini 2026βs analysis, 7 of 13 Brain-cell-types significant for ME/CFS were striatal (MSN) cell types from Dropviz. Provides the strongest mouse-based replication of the MSN/ME/CFS enrichment signal. Quality:: Very High (Cell, landmark reference atlas, highly cited, peer-reviewed) Certainty:: 0.92 Limitations:: Mouse model β cross-species gene mapping introduces noise and false positives; limited to 9 brain regions; cell-type naming ambiguities (GP = striatal); mouse brain cell types do not perfectly map to human.
6 Siletti et al. 2023 β Human Brain Cell Atlas (Reference Resource)
Key Findings::
- Comprehensive single-nucleus RNA-seq atlas of entire human brain: 3.3 million cells, 105 dissections
- 31 superclusters, 461 clusters, 3,313 subclusters of transcriptionally defined cell types
- High neuronal diversity outside cortex β hypothalamus, midbrain, hindbrain represented
- eMSN (eccentric medium spiny neuron) cluster identified: GAD1/GAD2 (CXCL14+ DRD1+ ADARB2+)
- Publicly available resource used by Duncan 2025 pipeline
Conclusion:: Landmark reference atlas confirms eMSN as a distinct human brain cell type. The resolution needed (461 clusters) to test cell-type-specific enrichment for ME/CFS. Relevance:: The Duncan 2025 pipeline β applied to ME/CFS by mecfsscience.org β uses the Siletti 2023 atlas as input. eMSN is the top cell-type hit for ME/CFS from this atlas. Without this atlas, eMSN-level resolution would not be achievable from GWAS data. Quality:: Very High (Science, 3.3M cells, landmark study, peer-reviewed) Certainty:: 0.95 Limitations:: Postmortem tissue quality; only 3 donors (limited genetic diversity); cell-type clustering is algorithmic β boundaries approximate; does not address ME/CFS or any disease associations.
7 Duncan et al. 2025 β Human Brain Atlas Cell-Type Enrichment Pipeline
Key Findings::
- Applied MAGMA cell-type enrichment to 18 brain-related traits using Siletti 2023 Human Brain Atlas (461 cell types)
- Schizophrenia enriched in cortical layer 2/3 excitatory neurons, inhibitory interneurons, MSNs
- Depression enriched in cortical layer 5/6 excitatory neurons
- Alcohol consumption and sleep duration enriched in medium spiny neurons
- Published code: github.com/Integrative-Mental-Health-Lab/linking_cell_types_to_brain_phenotypes
- Established h2-MAGMA method for partitioning heritability by cell type
- Pipeline used by mecfsscience.org for their ME/CFS eMSN enrichment result (analysis 5)
Conclusion:: The Duncan 2025 pipeline provides the most systematically validated framework for mapping GWAS signals to brain cell types. Its application to ME/CFS identified eMSN as the top cell-type hit. Relevance:: mecfsscience.org analysis 5 directly uses this pipeline with DecodeME+MVP summary statistics. The eMSN finding for ME/CFS comes from this specific methodology. Duncan et al. also show that MSN enrichment is NOT unique to ME/CFS (found for sleep, alcohol, schizophrenia) β interpretation must account for this. Quality:: Very High (Nature Neuroscience, peer-reviewed, publicly available code, validated on 18 traits) Certainty:: 0.88 Limitations:: Pipeline engineered for psychiatric/brain phenotypes β validation in non-psychiatric conditions is indirect; HBA has only 3 donors; MAGMA window (30/10kb) may miss distal regulatory elements; fine cell-type specificity remains method-dependent (Brouwer 2026 review).
8 GTEx Consortium 2020 β Tissue Gene Expression and eQTL Atlas
Key Findings::
- Multi-tissue gene expression and eQTL atlas from 838 postmortem donors across 49 tissues
- 15,201 RNA-seq samples total; all 13 brain regions profiled
- GTEx v8 (2019 release) used by DecodeME preprint for MAGMA tissue enrichment analysis
- All brain regions show significant enrichment for ME/CFS-associated genes
- Pre-processed tissue expression data available via FUMA platform for MAGMA input
Conclusion:: GTEx is the primary tissue-level expression reference for GWAS enrichment. The uniform ME/CFS enrichment across all brain regions suggests broad neural involvement rather than region specificity. Relevance:: DecodeMEβs MAGMA tissue enrichment, which first identified CNS enrichment for ME/CFS, used GTEx v8 as the expression reference. The finding that ALL brain regions are significant (not just striatum) constrains interpretation: tissue-level analysis cannot resolve brain-region specificity. Quality:: Very High (Science, 838 donors, landmark resource, peer-reviewed) Certainty:: 0.90 Limitations:: Postmortem tissue; steady-state mRNA proxy (not protein/function); limited donors per tissue type; tissue-level resolution only β cannot distinguish brain regions or cell types; expression may differ between living and postmortem tissue.
9 Lee et al. 2026 β Genetic Overlap ME/CFS-IBS-Psychiatric Traits (DESCARTES Enrichment)
Key Findings::
- Cross-trait genetic correlation analysis (LDSC): ME/CFS shares modest genetic correlation with IBS and psychiatric traits
- Applied DESCARTES Human Fetal Atlas for cell-type enrichment of ME/CFS GWAS
- Inhibitory interneurons showed significant enrichment in fetal brain samples
- Third independent cell atlas replicating CNS enrichment for ME/CFS (after Dropviz and HBA)
- Consistent neuronal rather than immune cell-type involvement across all three atlases
Conclusion:: Independent replication of CNS enrichment using a third atlas (DESCARTES). The specific cell-type hit (inhibitory interneurons vs eMSN) differs, highlighting that fine cell-type specificity is atlas/method-dependent while the broad neuronal signal is robust. Relevance:: Important for triangulation: 3 different atlases (Dropviz, HBA, DESCARTES) all point to CNS/neuronal enrichment for ME/CFS. The specific cell type varies by atlas β this should be communicated as uncertainty, not contradiction. Quality:: Medium (preprint, not peer-reviewed, but methods are standard) Certainty:: 0.45 Limitations:: Preprint (not peer-reviewed); DESCARTES uses fetal samples β may not represent adult brain cell types; cross-trait analysis exploratory; specific interneuron finding may be atlas-dependent.
10 Snyder et al. 2025 β Rare Variants Implicating Neuronal Genes in ME/CFS
Key Findings::
- Whole exome sequencing of ME/CFS patients (Stanford Medicine)
- Enrichment of rare damaging variants in neuronal genes
- Synaptic communication and neural development gene ontologies significantly enriched
- No immune gene enrichment from rare variant analysis
- Converges with DecodeME common variant findings: both rare AND common variation point to synaptic/neuronal biology
Conclusion:: Rare and common variants converge on neuronal/synaptic biology in ME/CFS. Both ends of the allele frequency spectrum point to the same biological domain β a strong triangulation signal. Relevance:: Provides orthogonal evidence (rare variant burden, not GWAS enrichment) that neuronal dysfunction β not immune dysfunction β is the genetically encoded risk in ME/CFS. Strengthens the DecodeME enrichment findings with a completely different methodology. Quality:: Medium (preprint, not peer-reviewed, Stanford group reputable) Certainty:: 0.45 Limitations:: Preprint (not peer-reviewed); exact sample size not confirmed; rare variant analysis has inherently lower statistical power than common variant GWAS; exact sample overlap with DecodeME participants unknown.
11 Chaudhuri & Behan 2000 β Fatigue and Basal Ganglia (Striatal-Thalamic Hypothesis)
Key Findings::
- Introduced concept of central fatigue (distinct from peripheral neuromuscular fatigue) in neurological diseases
- Proposed that central fatigue results from failure in integration of limbic input and motor functions within basal ganglia
- Specific mechanism: dysfunction in the striatal-thalamic-frontal cortical system
- Supported by neuropathological data in Parkinson's disease and multiple sclerosis
- Central fatigue in ME/CFS, MS, and Parkinson's posited to share common basal ganglia circuit pathology
- Distinguished physical fatigue from mental fatigue β both proposed as central in origin
Conclusion:: The striatal-thalamic-frontal cortical system hypothesis predicts that ME/CFS pathology should map to striatal cell types. The 2026 cell-type enrichment results (eMSN in striatum) provide genetic evidence consistent with this 25-year-old hypothesis. Relevance:: This is the theoretical framework that the mecfsscience.org enrichment results appear to support. Proposed in 2000, predating GWAS, single-cell sequencing, and enrichment methods β notable as a hypothesis confirmed by modern tools rather than contradicted. Quality:: Medium (review article, hypothesis-level, no experimental validation at the time, peer-reviewed in J Neurol Sci) Certainty:: 0.55 Limitations:: Hypothesis paper with no original data; mechanistic pathways speculative; predates modern genetic methods; prediction about basal ganglia involvement is now supported by enrichment data but the specific circuit mechanism remains unproven; the full striatal-thalamic-frontal cortical pathway has not been directly tested in ME/CFS.