Aggregated Evidence

 Quick Analysis Plan Copied

 Long Analysis Plan

Key evidence constraints

No record supplies B05 and 154Q/2Q outcomes measured in the same experiment, so a true meta-analytic effect estimate cannot be computed from these records alone. What exists: (1) the Sca1 154Q/2Q knock-in model with quantitative endpoints — impaired Morris water maze latency (p = 0.000001), reduced CA1 LTP at 24 weeks (WT 163 ± 7.4 vs mutant 137 ± 9.1, p = 0.026), ~50% lower Purkinje dendritic fluorescence, and region-specific mutant ataxin-1 solubility; Sca1 154Q/2Q knock-in characterization (2) B05 transgenic-line therapy studies — TRTK12 (anti-S100B) improving rotarod performance and dendritic morphology TRTK12 therapy in B05 mice and ASIC1a knockdown improving motor function and Purkinje survival ASIC1a knockdown in SCA1 mice; and (3) a modern 146Q/2Q microglia-deletion (mKO) dataset with rotarod at 14/24/34 weeks, Barnes maze, and fear conditioning ATXN1 microglial mKO therapy outcomes.

Graph reports only the sourced LTP values; no B05 LTP data exist to plot alongside.

Analysis plan

Step 1 — Effect extraction: parse all included papers with evidence-unit schemas; for B05 (TRTK12, ASIC1a) extract standardized rotarod latency differences vs vehicle/littermate controls; for 154Q/2Q extract MWM latency, LTP, dendritic fluorescence; for 146Q/2Q mKO extract rotarod (14/24/34 wk), Barnes maze cognitive score, fear conditioning freezing (%), and DEG counts (426 human microglia, 968 cerebellar). Step 2 — Harmonization: since B05 is an overexpression transgenic line and 154Q/2Q a knock-in, model type must be a stratifying covariate, not pooled naively; convert each outcome to Hedges' g with reported means ± SD/SEM (SEM converted; n extracted from methods; where n is unreported, flag as missing rather than imputing). Step 3 — Meta-analysis: random-effects (DerSimonian-Laird or restricted maximum likelihood) per outcome family using Python (statsmodels, numpy, custom PALE-style variance weighting); leave-one-out sensitivity; funnel plot/Egger test for publication bias. Step 4 — Cross-species/cross-model annotation: record species, line, CAG length, age, and Cre-driver caveats (Lyve1-Cre is not microglia-exclusive Lyve1-Cre specificity limitation). Step 5 — Visualization: forest plots per endpoint; subgroup panels B05 vs knock-in.

Critical gaps and blind spots

Missing: raw B05 therapy effect sizes and n values, any B05 cognitive/physiology data, and any therapy tested in 154Q/2Q mice. The 154Q/2Q solubility gradient (cerebellum highest at 0.58) Regional ataxin-1 solubility values may itself confound therapy response comparisons across regions. Confidence in the plan is high; confidence in any pooled B05-vs-154Q/2Q effect estimate from current data is nil — direct comparative datasets are required to disprove the conclusion that models cannot yet be validly pooled.