Back

Human CSF proteogenomics links genetic variation to neurodegenerative disease proteins

Puerta, R.; Garcia-Gonzalez, P.; de Rojas, I.; Capdevila-Bayo, M.; Olive, C.; Munoz-Morales, A.; Bayon-Bujan, P.; Valenzuela, A.; Yang, C.; Timsina, J.; Liu, M.; Chakkarai, S.; Sotolongo-Grau, O.; Calm, B.; Miguel, A.; Solivar, A.; Montrreal, L.; Martinez, M.; Khan, A.; Zhao, F.; Tantinya, N.; Rosende-Roca, M.; Alegret, M.; Moreno-Grau, S.; Fernandez, M. V.; Marquie, M.; Valero, S.; Cavazos, J. E.; Sanz, P.; Montalban, X.; Tarraga, L.; Smets, B.; Boada, M.; Seshadri, S.; Sargurupremraj, M.; Cruchaga, C.; Cano, A.; Cabrera-Socorro, A.; Ruiz, A.

2026-02-22 neurology
10.64898/2026.02.12.26345733 medRxiv
Show abstract

The cerebrospinal fluid (CSF) proteome offers a direct readout of central nervous system (CNS) biology but its genetic architecture remains incompletely defined. We conducted the largest single-site CSF genome-wide association study (GWAS) to date, analysing 7,092 SomaScan proteins in 1,259 individuals. Using a covariate-adjusted model including proteomic PCs and disease status, we identified 1,971 genome-wide significant pQTLs (954 cis, 971 trans), 1,409 of which replicated in an independent CSF dataset. We discovered 264 previously unreported loci, replicated 511 associations, refined 80 known loci, and 265 proxy-based associations. Using a previously published reproducibility framework, we show that robust discovery concentrates in reliable measurements, underscoring the importance of rigorous quality control. Enrichment analyses revealed immune/complement and extracellular matrix biology. Mendelian randomization prioritised causal proteins: PILRA, TREM2, IL34, CR2, SHARPIN and ERBB1 (Alzheimers disease); BST1 and GPNMB (Parkinsons disease); STX6 (Creutzfeldt Jacobs disease); and ATXN3 and B4GALNT1 (Amyotrophic lateral sclerosis), providing a scalable framework for orthogonal target validation in neurodegeneration. HighlightsCSF A{beta}42 and p-tau, CSF total protein and Qalb are major contributors to the proteomic variance and may act as potential confounders. Most pQTLs were found in proteins classified as "reproducible" based on our CSF proteomic score. The most stringently adjusted GWAS model maximized pQTL discovery among the highest-reproducibility proteins (score 1 and A). We identified 264 novel CSF pQTLs that were not described in previous analyses, replicated 511 CSF pQTLs, 80 map refinements and 265 proxy SNPs, predominantly involved in immune-related, inflammatory, and extracellular matrix mechanisms. We found 281 CSF pQTLs that were also systemic modulators of plasma protein levels that were enriched in immune-related and extracellular matrix mechanisms. We identified and validated several causal proteins associated with AD and other neurodegenerative disorders.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.