Design Experiments From Reported Results

 Quick Analysis Plan Copied

 Long Analysis Plan

Evidence grounding and honest gaps

Supplied records establish that NSD2 loss is a strong chromatin-remodeling perturbation: NSD2 depletion reduces H3K36me2 on gene bodies while H3K27me3 increases at silenced loci, with hundreds of genes shifting expression in both directions NSD2 H3K36 methylation loss remodels gene expression, and in hair cells NSD2 loss opens chromatin at specific loci and upregulates 216 genes with ECM pathways enriched NSD2 loss opens chromatin and de-represses loci. Anti-PD1 efficacy in the human breast-cancer BioKey study tracks pre-treatment PD1+ T-cell fractions and clonal expansion of cytotoxic CD8 cells (AUC ~0.92–0.96), with MHC expression on cancer cells lower in non-expanders Single-cell anti-PD1 response map in breast cancer. NSD2 is a validated oncogenic dependency with emerging chemical tools: a catalytic-relevant E3-recruitment degrader (benzaldehyde-based) reduces NSD2 protein ~75% at 1 µM, DC50 ≈ 460 nM NSD2 degrader chemistry and structural basis, and NSD2 knockdown alone increases apoptosis and cisplatin sensitivity in osteosarcoma xenografts NSD2 knockdown sensitizes tumors in vivo. Key gap: no supplied record directly demonstrates that NSD2 catalytic inhibition de-represses transposable elements or triggers a viral-mimicry/interferon response in NSD2-amplified tumors. The immune-sensitization arm is a well-motivated but currently untested inference; the design below must be powered to falsify it.

Experimental design (2×2×2 factorial, isogenic pairs)

Generate isogenic pairs: (a) endogenous NSD2-amplified line plus CRISPR-mediated amplification loss or catalytic-dead rescue, and (b) matched non-amplified line engineered for NSD2 overexpression; verify by WGS/CNV-qPCR, H3K36me2 Western/CUT&Tag, and growth-rate matching to control for proliferation confounds. Use a selective NSD2 catalytic inhibitor (or FBXO22-recruiting degrader as orthogonal chemical validation) at an established non-cytotoxic dose; include anti-PD1 vs isotype. Arms per line: vehicle, NSDi, anti-PD1, NSDi+anti-PD1 (8 groups minimum, n = 8–10 mice/group based on prior anti-PD1 murine power in E0771 designs where n ≈ 9–10/group detected p ≈ 0.02 effects Murine anti-PD1 power and diet confound warning). Critical confound control: hold diet constant (soy-free/isoflavone-free chow), since isoflavones can ablate anti-PD1 responses in murine mammary models.

Primary outcomes and readouts

Tumor control: longitudinal caliper volume and time-to-progression; test for synergy (Bliss or two-way ANOVA interaction term). Molecular (tumors harvested D7 and D14 post-treatment): bulk RNA-seq (TE-counting pipeline: STARsolo/TEtranscripts with RepeatMasker annotations) for LTR/LINE/ERV expression; IFN-signature scoring via Hallmark/ISG15/OAS gene sets; Nanopore or short-read ATAC-seq on matched tumors to confirm NSD2-loss-dependent chromatin opening at TE loci. Immune: flow cytometry (CD8, GZMB, TIM3, PD1), IHC spatial mapping of CD8 density, and scRNA-seq + TCR-seq on a subset to check whether NSDi expands the PD1+ cytotoxic clonotype pool that the BioKey data links to response. Bioinformatics pipeline: Snakemake/nf-core RNA-seq → TEtranscripts → DESeq2 (line × treatment interaction) → GSEA (MSigDB Hallmark IFN, GO interferon-gamma response) → correlation of TE expression with CD8 infiltration (Spearman) → immune deconvolution (xCell/MuSiC) → scRNA analysis in scanpy (T-EX annotation, STARTRAC clonotype expansion). Pre-register the falsification criterion: NSDi must de-repress TEs and raise IFN scores in NSD2-amplified tumors only; if amplification status does not interact with treatment, the sensitization hypothesis is falsified.

Above: illustrative expected pattern only (not measured data) — the experiment is designed to detect whether the combination arm (red) exceeds additive suppression; if it does not, the hypothesis fails.

Blindspots

(1) NSD2 loss effects are documented for ECM/adipogenic loci, not TEs — this is the pivotal unverified premise. (2) Catalytic inhibitors vs degraders differ: degrader chemistry carries aldehyde warhead liabilities and lacks in vivo PK data. (3) Mouse strain and diet modulate anti-PD1 strongly; isoflavone-containing chow could mask or mimic sensitization. (4) Amplification engineering may create passenger-locus artifacts; use two independently derived clones per direction. (5) Cell-line-derived xenografts under-represent tumor heterogeneity; a syngeneic orthotopic confirmation is advisable.