APOEε4 Multi-Omic Project

Lead author: Yaroslav Markov
Primary affiliation (at time of work): Yale University

This repository contains code, documentation, and analysis scipts for a multi-omic study of APOEε4-associated proteomic and DNA methylation changes in Alzheimer disease using post-mortem human brain tissue.

---

Project Motivation

Alzheimer disease (AD) is the leading cause of dementia worldwide and a major contributor to morbidity, mortality, and healthcare burden. The APOEε4 allele is the strongest common genetic risk factor for late-onset AD, but how APOEε4 shapes molecular changes in the brain—especially which changes are early, potentially compensatory versus late and overtly pathological—remains unclear.

This project addresses this using a uniquely APOEε4-enriched subset of ROSMAP, enabling us to dissect potential APOEε4-related risk and resilience mechanisms at the molecular level.

---

Study Overview

We analyzed prefrontal cortex (BA10) tissue from a subset of the Religious Orders Study and Rush Memory and Aging Project (ROSMAP):

• Proteomics (DIA LC–MS): n = 302
• DNA methylation (EPIC array): n = 310
• Genetic data / PRS (imputed WGS/WGS-derived genotypes): n = 254

Key cohort feature:

• 69% (n = 214) were APOEε4 carriers
• 19.6% of carriers (n = 42) had no AD pathology by NIA–Reagan criteria

We model:

• APOEε4 × AD interactions at the protein and CpG level
• Mediation of APOEε4 → AD by selected proteins and CpG sites
• Associations with neuropathology (CERAD, Braak, NIA-Reagan) and polygenic risk scores (PRS)
• Predictive performance of proteomic / epigenetic / PRS features stratified by APOEε4 status

---

Core Findings

1. Biphasic model of APOEε4-driven pathology

In the absence of AD pathology, APOEε4 carriers show lower levels of synaptic and metabolic proteins (e.g., VAMP1, SYN3, CASKIN1, GLUD1, PI4KA), consistent with early synaptic and metabolic vulnerability.

In APOEε4 carriers with AD, we instead see upregulation of inflammatory and proteostatic proteins, including:

• GNAO1, AHNAK, FGG, HEBP1, APEX1, RAB4A, SLC12A5, LRP1, BAG6
• Hypermethylation of cg06329447 in ELAVL4

Together, these support a biphasic model:
early synaptic/metabolic disruptions → later inflammatory / proteostasis and epigenetic changes in APOEε4-associated AD.

2. APOEε4 × AD interaction and mediation

• Several proteins and CpG sites show significant APOEε4 × AD interactions, indicating genotype-specific molecular responses to pathology.
• Mediation analyses identify:
  • Proteins such as GRIPAP1, GSTK1, VAMP1, CASKIN1, DPP3, SYN3, FGG mediating a substantial portion (~9–33%) of APOEε4’s effect on AD.
  • Hypermethylation at cg06329447 in ELAVL4 mediating ~12% of the APOEε4 effect.

3. PRS: APOE-specific vs genome-wide risk

We computed two polygenic risk scores (PRS):

• Non-APOE PRS: genome-wide, excluding the APOE region
• APOE-inclusive PRS: genome-wide, including APOE

Key observations:

• The non-APOE PRS is strongly associated with AD neuropathology but not with the core APOEε4-related proteins or CpGs we identify.
• The APOE-inclusive PRS is significantly associated with eight AD-related proteins in APOEε4 carriers, supporting that these markers are tied specifically to APOEε4-related genetic risk, not generalized AD polygenic risk outside APOE.

4. Predictive modeling stratified by APOEε4

In subjects with complete data for proteomics, DNAme, and PRS:

• In non-carriers, PRS performs best for AD classification (AUC ≈ 0.73).
• In carriers, proteomic and DNA methylation markers outperform PRS (AUC up to ≈ 0.74).

---

Data locations (for Albert Higgins-Chen lab)

Note: These data are not included in this public repository. The paths below are internal pointers for authorized Higgins-Chen lab members.

1. DNA methylation (EPIC array)
   Google Drive:
   4_Research Resources -> Datasets -> ROSMAP -> Methylation

2. Proteomics (raw + processed)
   Google Drive:
   4_Research Resources -> Datasets -> ROSMAP -> Proteomics -> Proteomics MultiRegion

3. Whole-exome sequencing (WES)

   • Raw FASTQs: download from Yale storage tape using:
     WES_pipeline/00_download_fastq.sh
   • Processed WES + related files: Google Drive:
     4_Research Resources -> Datasets -> ROSMAP -> WES

4. Polygenic risk scores (PRS)
   Google Drive:
   4_Research Resources -> Datasets -> ROSMAP -> PRS

---

Repository Structure

.
├── README.md                # Project overview and entry point
├── docs/                    # High-level project and methods documentation
│   ├── README.md            # Overview, methods, and FAQs
│   ├── data_overview.md     # Data types, cohorts, and key variables
│   └── methods_summary.md   # Condensed methods description
├── notebooks/
│   ├── 01_pheno_exploration_and_heatmaps.R
│   ├── 02_epigenetic_biomarker_associations.R
│   ├── 03_ewas_AD_status.R
│   ├── 04_wgcna_DNAm_modules.R
│   ├── 10_proteomics_import_and_cleaning.R
│   ├── 11_proteomics_normalization_and_integration.R
│   ├── 12_proteomics_ruv_normalization_and_differential.R
│   ├── 20_methylation_interactions_and_mediation.R
│   ├── 21_protein_interactions_and_mediation.R
│   ├── 30_prs_associations_and_modeling.R
│   └── 40_multiomic_integration_and_final_analyses.R
├── WES_pipeline/
│   ├── config/
│   │   ├── hg19_idt_target_regions.bed
│   │   ├── hg19_idt_and_spikein_regions.bed
│   │   ├── hg19_spikein_regions.bed
│   │   ├── hg38_idt_and_spikein_regions.bed
│   │   ├── hg38_idt_target_regions.bed
│   │   └── hg38_spikein_regions.bed
│   ├── 00_download_fastq.sh
│   ├── 01_fix_sample_names_and_paths.sh
│   ├── 02_merge_lanes_fastq.sh
│   ├── 03_quip_to_gz.sh
│   ├── 04_fastqc_raw.sh
│   ├── 05_multiqc_raw_QC.sh
│   ├── 06_trimgalore.sh
│   ├── 07_build_decoy_reference.sh
│   ├── 08_bwa_align_and_target_filter.sh
│   ├── 09_sort_bam_with_picard.sh
│   ├── 10_merge_lanes_bam_with_picard.sh
│   ├── 11_mark_duplicates.sh
│   ├── 12_collect_alignment_metrics.sh
│   ├── 13_base_recalibration.sh
│   ├── 14_apply_bqsr.sh
│   ├── 15_get_pileups_and_contamination.sh
│   ├── 16_mutect2_calling.sh
│   ├── 17_learn_read_orientation_and_sort_vcf_and_filter_mutect_calls.sh
│   ├── 18_filter_variants_keep_pass.sh
│   ├── 19_snpeff_annotation.sh
│   ├── 20_snpsift_annotate_dbsnp_clinvar_gwas.sh
│   ├── 21_snpsift_genesets.sh
│   ├── 22_bgzip_and_tabix_vcfs.sh
│   ├── 23_merge_sample_vcfs.sh
│   └── readme_WES_pipeline.md