450K arrays in the internal duplicate study
Karam computational research · Updated 24 October 2026
ClockQC: measure the measurement.
How much of an aging-clock change comes from technical variation, preprocessing or the clock itself? Our public-data analyses test those questions before drawing biological conclusions.
Recorded duplicate pairs retained in the primary analysis
Clock versions across five original/PC families
Research status: completed computational analyses of public datasets. These are Karam reanalyses of other investigators’ measurements, not new biological experiments or clinical validation. The reports have not been presented here as peer-reviewed publications.
Latest result · v0.10
A plausible explanation did not hold up.
Adding 65 SNP probes did not resolve the small mismatch with the deposited matrix. Beta-value reconstruction RMSE increased by 0.73%; PCHorvath1 duplicate disagreement was essentially unchanged. Exact historical preprocessing remains unresolved.
What the broader experiment tells us
Preprocessing changes the result.
In v0.9, the adapted P2Q pipeline reduced PCHorvath1 duplicate disagreement by 22%, while between-donor score spread fell 29%. Lower disagreement alone did not establish improved measurement.
PC gains vary by clock.
In the separate four-donor EPIC study, PCPhenoAge and PCDNAmTL had 61% and 73% lower within-donor SD among independent preparations. These are dataset-specific results, not a universal PC-clock noise floor.
Function is a separate endpoint.
These data do not measure day-to-day biological variation, clinically meaningful change, treatment response or aging reversal. A clock result must be evaluated alongside the function, identity and safety outcomes relevant to the study.
The external study has only four donors and shared slides; its intervals are descriptive. Original Hannum probe coverage failed the frozen primary eligibility threshold. See the full v0.9 report for exclusions, imputation, conditional intervals and signal-compression checks.
Keep the evidence trail visible
The v0.10 note includes the hypothesis, fixed comparison, complete-array result and verification limits. Its downloads distinguish an independent score-table audit from the larger extension package that requires the v0.9 environment.