# Skeptical Commentary on Cell-Type Specific Expression Patterns in SEA-AD
I must press on several methodological vulnerabilities that deserve scrutiny before accepting these cell-type specific conclusions. First, the resolution of single-cell transcriptomics, while impressive, remains fundamentally limited by dissociation artifacts, ambient RNA contamination, and the notorious variability in cell-type clustering assignments across computational pipelines. How robustly do the reported expression patterns replicate across different clustering algorithms, and critically, have the authors validated their cell-type identities against established protein-level markers using spatial transcriptomics or immunofluorescence on the *same tissue samples*? Without this orthogonal validation, we're essentially reading the tea leaves of bioinformatic assumptions. The SEA-AD dataset, while valuable, pools samples across different post-mortem intervals, neuropathological stages, and potential comorbidities—confounding variables that could easily produce spurious cell-type associations rather than genuine disease mechanisms.
Second, I'm concerned about the inference of causality from correlation. Demonstrating that neurodegeneration-associated genes show preferential expression in specific cell types tells us remarkably little about whether those cells are drivers, passengers, or collateral damage in disease progression. The authors must contend with a fundamental problem: Are microglia upregulating neuroinflammatory genes *because* they're responding to neuronal dysfunction, or does their expression pattern contribute to pathology? Without manipulative studies—selective knockdown or overexpression in defined cell populations—we cannot distinguish these scenarios. Furthermore, I notice the discussion largely overlooks the substantial inter-individual heterogeneity documented within the SEA-AD cohort itself; cherry-picking consistent patterns across samples risks obscuring the biological reality that neurodegeneration may engage fundamentally different cellular players in different individuals.
Finally, the translational leap from expression pattern to therapeutic target remains premature. Many of these genes show cell-type enrichment in normal aging brains as well—a critical baseline comparison that appears underexplored. Until the authors demonstrate that their identified cell-type specific patterns are *specific to disease* rather than normal senescence, or that perturbing these expression patterns in model systems actually ameliorates pathology, we should resist the temptation to declare these findings disease mechanisms worthy of drug development attention.