POST-DEFENSE FOLLOW-UPS · JUNE 16 2026

Committee questions, answered.

Doxorubicin cardiotoxicity in stem-cell cardiomyocytes · follow-up analyses requested at the proposal defense · Waleed Osman, George Mason University.

✓ Proposal defense passed — committee voted in favor

follow-up analyses

1.3×10⁻⁴

anthra > TKI (1 µM)

4/4

iron trend tests significant

OXPHOS

drives the EP collapse

D-1 · Dr. Sobie

The response is shared across anthracyclines — and distinct from other drug classes

Each of the 54 DToxS drugs correlated with the corrected hESC signature, grouped by class.
At 1 µM: anthracycline mean ρ 0.377 vs TKI −0.016 vs other −0.026.
anthracycline > TKI p = 1.3×10⁻⁴; anthra > other p = 5.6×10⁻⁵; Kruskal–Wallis p = 4.1×10⁻³.
Shared across all four: EPI #1, DOX #2, DAU #3, IDA #6 of 54 — a class property, not a doxorubicin quirk.
TKI ≈ other (p = 0.45) — it is specifically the anthracycline class that stands out.

The class separation is dose-specific — sharp at 1 µM, weaker at 2.5 µM (anthra>TKI p = 0.05), because the high-dose signature is broader and more global.

cross-class comparison — Per-drug Spearman correlation with the corrected hESC dox signature (1 µM), by class. Anthracyclines (n=4) sit well above TKIs (n=22) and other drugs (n=28), which cluster near zero.

D-2 · Dr. Jafri

OXPHOS / ROS changes — not ion-channel expression — drive the high-dose EP collapse

Dose	Baseline	Conductance	OXPHOS/ROS	Combined
1 µM	339.6	347.6 (+2.4%)	246.6 (−27%)	252.5 (−26%)
2.5 µM	339.6	EAD/fail	237.4 (−30%)	EAD/fail

Answers the committee's request to show the conductance vs OXPHOS contributions individually and combined — computed in the 11-scenario simulation, now surfaced explicitly.

The two axes push opposite directions: conductance prolongs APD, OXPHOS/ROS shortens it.
The bioenergetic axis dominates the combined response.
At 2.5 µM the combined model collapses into an early afterdepolarization / failure.

EP decomposition — APDₐ₀ under each axis in isolation and combined (Paci 2013, Shim scaling). Conductance changes mildly prolong; OXPHOS/ROS shortens and dominates; 2.5 µM combined destabilizes into EAD/failure.

D-3 · Dr. Jafri / Dr. Hansen

The modest iron-handling rise is statistically robust

Significance is the dose-trend across replicates, not the size of the mean shift (2.74 → 2.90).
Four concordant tests on the full 19-gene module: Spearman p = 1.2×10⁻⁴, linear regression p = 3.3×10⁻⁴, Kruskal–Wallis p = 1.2×10⁻², Mann–Whitney 0-vs-2.5 µM p = 1.4×10⁻².
Gene level: import/storage up (FTL, TFRC, FTH1, TF); export/regulation down (HMOX1, FXN) — the iron-overload-response pattern.
Restricting to a 5-gene subset is unstable (ssGSEA on few genes mixes the up and down arms) — the curated a-priori module is the right unit; per-gene log₂FC is the readout for a subset.

Iron dysregulation, not ferroptotic death — rising import with stressed storage capacity is a response to iron overload.

iron significance — (A) Full 19-gene iron module per replicate (mean ± SD) with the four trend tests. (B) Per-gene log₂FC at 2.5 µM — import/storage up, export/regulation down.

D-4 · Dr. Hansen

The on-slide iron p-value — sourced and confirmed

The slide's “ρ = 0.89, p = 1×10⁻⁴” is the Spearman correlation of per-sample iron-handling ssGSEA score vs dose.
Exact values: ρ = 0.887, p = 1.198×10⁻⁴.
Now backed by three additional concordant tests (D-3 above).

All analyses run on the GMU Hopper HPC from the corrected reanalysis. Full write-up: validation/new_analyses/POST_DEFENSE_FOLLOWUP.md · scripts d1_crossclass.py, d3_ssgsea_sig.py. © 2026 DoxToxOmics · George Mason University.