The proteome atlas framework represents a paradigm shift in how neurodegenerative diseases are classified and understood. Rather than relying solely on clinical phenotype and histopathological hallmarks, proteome atlases map the complete landscape of protein abundance, post-translational modifications, solubility shifts, and network co-expression patterns across disease subtypes. This molecular approach reveals hidden biological signatures that transcend traditional clinical boundaries, enabling more precise patient stratification, earlier diagnosis, and targeted therapeutic development.
Neurodegenerative diseases have historically been defined by their clinical presentation and post-mortem pathology. Alzheimer's Disease is diagnosed by amyloid plaques and tau tangles, Parkinson's Disease by alpha-synuclein Lewy bodies, and Frontotemporal Lobar Degeneration by TDP-43 or tau inclusions. However, this classification system fails to capture the considerable molecular heterogeneity within each disease category, the overlapping biological processes across diseases, and the preclinical stages where molecular changes precede clinical symptoms by years or decades.
The emergence of large-scale proteomics technologies — including mass spectrometry-based analysis of cerebrospinal fluid (CSF) and plasma, aptamer-based proteomics platforms (SomaScan, Olink), and proximity ligation assays — has enabled comprehensive protein quantification across thousands of analytes in patient cohorts. Proteome atlases integrate these data with network-based analyses (weighted gene co-expression network analysis, WGCNA) to identify disease-specific molecular signatures that can distinguish subtypes, predict progression, and highlight therapeutic targets.
Recent landmark studies have demonstrated the power of this approach in Frontotemporal Lobar Degeneration (Saloner 2025, Nat Neurosci), primary tauopathies (Kavanagh 2025, Acta Neuropathol), and Alzheimer's Disease (Bai 2020) [@saloner2025; @kavanagh2025; @bai2020].
The landmark 2025 Nature Neuroscience study by Saloner and colleagues analyzed cerebrospinal fluid proteomes from 116 carriers of autosomal dominant FTLD mutations compared with 39 non-carrier controls [1]. The study used aptamer-based proteomics (SomaScan v4) to quantify over 4,000 proteins across the cohort. Researchers applied weighted gene co-expression network analysis (WGCNA) to identify 31 protein co-expression modules, which were then correlated with both cross-sectional clinical indicators (CDR+NACC-FTLD, CSF neurofilament light chain, bilateral frontotemporal volume) and longitudinal cognitive trajectory measures.
This approach treats the CSF proteome as an integrated network rather than a collection of individual biomarkers, capturing the coordinated biological processes that characterize disease states.
The WGCNA analysis revealed that each major genetic subtype of FTLD harbors a distinct proteomic signature:
A striking finding was that all three genetic subtypes show decreased synaptic/neuronal modules and decreased autophagy modules, indicating convergent endpoints of neurodegeneration despite divergent upstream causes [1:1]. This convergence has profound therapeutic implications — interventions targeting synaptic protection or autophagy enhancement may be broadly applicable across FTLD subtypes.
The researchers validated their findings across independent cohorts, including the 4RTNI cohort (sporadic progressive supranuclear palsy-Richardson syndrome) and the BioFINDER 2 cohort (frontotemporal dementia spectrum clinical syndromes), using both SomaScan and Olink platforms [1:2]. This cross-platform replication confirms that the molecular signatures are robust and not artifacts of a single measurement technology.
The ability to validate findings in sporadic disease cohorts — not just genetic carriers — establishes the generalizability of proteome-based disease classification beyond the monogenic forms of FTLD.
The study identified "hub" proteins — highly connected nodes within the affected co-expression modules — as particularly promising for biomarker and therapeutic development [1:3]. These hub proteins represent the master regulators of disease-related biological processes. Network-based proteomics demonstrated potential for identifying replicable molecular pathways that could guide drug development for adults living with FTLD.
A parallel breakthrough came from the 2025 Acta Neuropathologica study by Kavanagh and colleagues, which applied sarkosyl fractionation and mass spectrometry to post-mortem brain tissue from three primary tauopathies: corticobasal degeneration (CBD), Pick's disease (PiD), and progressive supranuclear palsy (PSP) [3].
The study examined not just protein abundance, but also protein solubility — a critical distinction because disease-associated proteins often shift from soluble to insoluble compartments. The findings revealed that CBD and Pick's disease showed the greatest proteomic similarity in both the soluble and insoluble fractions, while PSP exhibited the most divergent profile [3:1]. This finding challenges the traditional grouping of these "4R tauopathies" and suggests that CBD and Pick's disease share more biological overlap than CBD and PSP.
The study identified consistent solubility changes across the tauopathies affecting four major biological categories:
Lysosomal regulators: Unique lysosomal proteins become more insoluble in distinct tauopathies, suggesting that lysosomal dysfunction plays a disease-specific role in each condition. SORT1 (sortilin) was identified as highly insoluble in CBD and aggregates to different extents across tauopathies [3:2].
Postsynaptic proteins: synaptic proteins showed altered solubility, consistent with the well-documented synaptic loss in these conditions and suggesting postsynaptic targeting as a therapeutic strategy.
Extracellular matrix (ECM): ECM proteins displayed distinct solubility patterns across subtypes, with MAPT carriers in the FTLD study showing particularly elevated ECM modules [1:4].
Mitochondrial proteins: Mitochondrial dysfunction is reflected in solubility shifts of mitochondrial proteins, consistent with the established role of energy metabolism disruption in neurodegeneration.
The Kavanagh study identified several proteins as promising biomarker candidates [3:3]:
Complementing the Kavanagh study, Morderer and colleagues (2025, Brain) used probe-dependent proximity profiling (ProPPr) to map the protein interaction landscape of phospho-tau across tauopathies [4]. ProPPr uses engineered peroxidase enzymes to label proteins in proximity to a bait protein (phospho-tau) in fixed tissue. This approach uncovered both similarities and differences in the phospho-tau-associated proteome between CBD, PSP, and Alzheimer's Disease.
The spatial resolution of ProPPr revealed that some tau-associated proteins cluster in specific subcellular compartments in disease-specific patterns, while others show more widespread changes. This spatial dimension of proteomic analysis adds a new layer of information beyond simple abundance or solubility measurements.
Large-scale proteomics has also transformed understanding of Alzheimer's Disease. The landmark study by Bai and colleagues (2020) performed deep proteomic profiling of human Alzheimer's Disease brain tissue, identifying over 10,000 proteins and revealing coordinated changes in specific biological pathways [5]. The study found that synaptic proteins, mitochondrial proteins, and proteins involved in proteostasis showed the most pronounced changes in Alzheimer's Disease, providing a molecular readout of the processes that underlie cognitive decline.
Wingo and colleagues (2021, Nat Neurosci) applied WGCNA to large-scale CSF proteomics in Alzheimer's Disease, identifying proteomic signatures of neuronal dysfunction that correlated with established biomarkers (amyloid-beta 42, phospho-tau, neurofilament light) and clinical outcomes.
Hammerling and colleagues (2025) analyzed CSF proteome network topology to identify Alzheimer's Disease subtypes with distinct biological signatures. The study found that network-based analysis outperformed traditional biomarker-based classification in predicting clinical progression, suggesting that the coordinated biological processes captured by proteomic networks carry more prognostic information than individual biomarkers.
Large-scale plasma proteomics has emerged as a complementary approach to CSF analysis, offering the advantage of accessible sampling. These studies demonstrate that plasma protein signatures can classify neurodegenerative disease subtypes with high accuracy. Plasma proteomics also shows utility for neurodegenerative disease classification and progression monitoring.
The convergence of findings across diseases reveals a molecular classification framework that transcends traditional clinical boundaries. The following diagram illustrates how proteome atlases classify neurodegenerative diseases by their molecular signatures:
The molecular signatures of Alzheimer's Disease subtypes include:
The proteomic signatures for Frontotemporal Lobar Degeneration subtypes are now well-characterized:
Emerging proteomic studies of Parkinson's Disease and related synucleinopathies have identified:
The proteomic signatures at the ALS-FTLD interface reveal:
The network-based approach to proteomics identifies "hub proteins" — highly connected nodes within disease-associated co-expression modules — as particularly valuable biomarkers. These hub proteins represent master regulators of disease biology, making them more likely to reflect the integrated state of the biological system than individual markers.
Key hub proteins identified across studies include:
Proteome atlases enable the integration of multiple fluid biomarkers into composite scores that capture disease biology more comprehensively than single markers. The AT(N) framework for Alzheimer's Disease — based on amyloid (A), tau (T), and neurodegeneration (N) biomarkers — represents an early implementation of this approach. Proteome atlases extend this framework by identifying additional biological axes (lysosomal function, synaptic integrity, inflammatory state, vascular function) that can be captured by fluid biomarkers.
Proteome atlases accelerate treatment discovery through several mechanisms:
Molecular subtype identification: By classifying diseases by their molecular signatures rather than clinical phenotype, proteome atlases enable more precise patient stratification for clinical trials. This allows enrichment of trial populations with patients most likely to respond to mechanism-specific interventions.
Convergence point identification: Despite the diversity of upstream causes (genetic mutations, environmental factors, aging), proteome atlases reveal convergent biological processes (synaptic loss, autophagy failure, neuroinflammation) that represent shared therapeutic targets applicable across multiple diseases and subtypes.
Network-based target prioritization: Hub proteins within disease-associated networks represent high-value therapeutic targets because modulating them is likely to affect multiple downstream processes. The proteome atlas approach identifies which hub proteins are most central to disease biology in each subtype.
Biomarker-driven trial design: Proteomic biomarkers enable monitoring of target engagement and disease modification in clinical trials, allowing more efficient go/no-go decisions and adaptive trial designs.
The molecular signatures identified by proteome atlases suggest several therapeutic strategies:
The next frontier in neurodegenerative disease proteomics is the integration of disease-specific atlases into a unified multi-disease framework. Such an integration would enable:
The proteome atlas framework connects to numerous related pages across the wiki:
The proteome atlas approach is rapidly evolving with several key developments on the horizon:
Single-cell proteomics: Technologies that enable proteomic profiling at single-cell resolution will reveal cell-type-specific molecular signatures within the brain, adding a new dimension to atlas-based classification.
Spatial proteomics: Techniques like ProPPr and imaging mass cytometry enable spatial mapping of protein expression, revealing how molecular signatures vary across brain regions and cell types.
Longitudinal proteomics: Repeated sampling of the same patients over time will enable mapping of proteomic trajectory changes, identifying critical windows for therapeutic intervention.
Integration with genomics: Combining proteomic data with genetic data will reveal how genetic variants affect protein abundance and function, enabling causal target identification.
Multi-omics integration: Integrating proteomics with transcriptomics, epigenomics, and metabolomics will provide a more complete picture of disease biology.
The proteome atlas represents a fundamental advance in the molecular understanding of neurodegenerative diseases, moving the field from clinical classification based on pathological hallmarks toward biology-driven classification based on measurable molecular signatures. This shift has profound implications for diagnosis, patient stratification, clinical trial design, and therapeutic development.
Saloner R, Staffaroni AM, Dammer EB, Johnson ECG, et al. Large-scale network analysis of the cerebrospinal fluid proteome identifies molecular signatures of frontotemporal lobar degeneration. Nat Neurosci. 2025. ↩︎ ↩︎ ↩︎ ↩︎ ↩︎
Soto C, Abel N. Proteostasis disruption in neurodegeneration. Nat Rev Neurosci. 2023. ↩︎
Kavanagh T, Balcomb K, Trgovcevic S, Nementzik L, et al. Differences in the soluble and insoluble proteome between primary tauopathies. Acta Neuropathol. 2025. ↩︎ ↩︎ ↩︎ ↩︎
Morderer D, Wren MC, Liu F, Kouri N, et al. Probe-dependent Proximity Profiling (ProPPr) Uncovers Similarities and Differences in Phospho-Tau-Associated Proteomes Between Tauopathies. Brain. 2025. ↩︎
Bai B, Wang X, Li Y, et al. Deep Multilayer Brain Proteomics of Human Alzheimer's Disease. Alzheimer's Dementia. 2020. ↩︎
Zetterberg H, Blennow K. NFLight in CSF and blood for neurodegeneration. Lancet Neurol. 2019. ↩︎