BGPT: Author Review: Leonard Petrucelli

Explore by Goal

Quick Explanation Copied

Leonard Petrucelli — scientific strength snapshot: The provided set of works strongly emphasizes mechanism-driven neurodegeneration biology (especially TDP-43/C9ORF72-related RNA/splicing/translation axes and proteostasis). Evidence strength appears highest when studies pair human tissue or patient-relevant models with quantitative omics and functional follow-up, but key general limitations recur across neurodegeneration work (cross-sectional designs, cell-type attribution limits, model-to-human translation uncertainty).

Long Explanation

Author Review: Leonard Petrucelli

Science-focused, evidence-weighted critique based only on the explicitly provided paper-level research data + the listed DOIs below.

1) What the provided evidence most strongly supports

Cryptic exon (CE) burden as a molecular stratification axis linked to proteome remodeling

The provided CE/hippocampus proteomics study reports that a CE burden–based subtyping strategy yields proteome-wide changes and stratifies samples in a way that is described as largely independent of classic AD pathology markers (ADNC and LATE-NC groups still show CE subtype structure, with CE burden correlating with depletion of CE-target–encoded neuronal proteins).
It reports a quantitative permanova effect size ("variance explained" value reported as 0.7).

TDP-43 level dysregulation can drive multi-system neurotoxic phenotypes in an in vivo transgenic model

The transgenic mouse study excerpt describes that human TDP-43 M337V overexpression driven by a prion promoter leads to dose-dependent reduction of endogenous mouse TDP-43, cytoplasmic/pTDP-43 aggregation and multiple downstream neurotoxic changes (gliosis, mitochondrial abnormalities, tau hyperphosphorylation) and early mortality patterns, and it emphasizes that phenotypes largely overlap WT TDP-43 overexpression, implying toxicity driven more by overall TDP-43 overexpression than mutation per se.

Multimodal cfDNA methylation signatures can distinguish ALS groups and correlate with progression readouts (as reported)

The provided cfDNA methylome excerpt reports a multimodal feature-selection + classification approach using MESA across four methylation-related modalities (CGI, DHS, nucleosome occupancy, and WPS), achieving LOOCV AUC 0.91 and independent-cohort AUC 0.85 for sALS vs controls, and reports additional correlations with a progression/biomarker readout (CSF pNfH correlations described for several promoter methylation markers).

2) Scientific strength: what looks rigorous vs what remains uncertain

Strength signals present in the provided excerpts

Cross-validation / external evaluation is explicitly mentioned in the cfDNA biomarker excerpt (LOOCV AUC reported and independent cohort AUC reported). That is a positive sign for avoiding overly optimistic single-sample splits, though it does not eliminate overfitting risk.
Human tissue + quantitative proteomics + perturbation consistency is described for the CE burden paper: postmortem proteomics plus an iPSC-derived neuron CRISPRi knockdown and detection of CE-linked proteomic directionality.

Primary uncertainties / potential failure modes explicitly highlighted by the excerpts

Cross-sectional human designs limit inference about causality and temporal ordering (CE burden and proteome remodeling could be correlated by shared upstream mechanisms rather than CE inclusion being causal).
Cell-type attribution limits: bulk proteomics and deconvolution approaches can obscure which cell types drive observed signals; iPSC neuron models lack full brain microenvironment.
Overfitting risk in high-dimensional multimodal models is a known vulnerability; the cfDNA excerpt itself flags small cohort size and overfitting risk.

3) Blind spots I would specifically try to falsify (using only the provided evidence context)

For the CE burden paper: test whether CE burden still predicts proteome shifts after controlling for multiple confounders (e.g., pTDP-43, phosphorylation states) beyond what is described; also test whether CE-target protein depletion remains after restricting analyses to matched pathology stages. The excerpt notes CE/pTDP-43 relationships but states they are not strictly co-dependent, leaving open a mechanistic heterogeneity question.
For the cfDNA biomarker paper: evaluate whether the same methylation signature holds across independent sites, different sequencing depths, and different cfDNA extraction/quality distributions; the excerpt flags targeted panel limitations and small cohort size as generalization constraints.
For the TDP-43 transgenic model: distinguish mutation-specific effects vs expression-level effects; the excerpt itself emphasizes overlapping phenotypes between M337V and WT TDP-43 overexpression, implying a dominant role for TDP-43 level. That creates a mechanistic blind spot if readers over-attribute to M337V biology.

4) Evidence-weighted bottom line (confidence-tagged)

Most supported claim from the provided evidence: The excerpted work collectively supports the idea that TDP-43-related molecular dysfunction can be linked to multi-level neurodegeneration biology (proteome changes and cellular phenotypes), and that epigenetic liquid-biopsy signatures may carry diagnostic/prognostic signal in ALS (with ML performance reported but generalization still a key uncertainty).

Confidence level: Moderate. Reason: the provided excerpts contain quantitative measurement + some validation/functional follow-up, but they also explicitly flag limitations typical for neurodegeneration (cross-sectional design, bulk attribution, cohort/generalization boundaries, and mutation-vs-expression ambiguity in animal models).

Feedback:

Updated: April 11, 2026