Definition
Correlation between digital twin harvest timing deviation and downstream processing failure. Isolates the cost that later harvests impose on cell viability, host cell protein contamination, and chromatography burden.
The Measurement Gap
When a digital twin consistently recommends harvesting later than experienced operators would choose, it chases marginally higher titer at the cost of lower cell viability at harvest. Lower viability means more cell debris, higher host cell protein contamination, and greater burden on downstream chromatography. But these downstream failures are attributed to normal process variation rather than to the twin's timing recommendations. Batch Success Rate doesn't track the correlation between harvest timing and subsequent downstream failures. Yield metrics show improvement because titer is higher at harvest. Contamination rates during purification are tracked but not correlated with harvest timing. The connection exists only in the operators' intuition, and intuition doesn't appear in a deviation database.
Decision Supported:
Is the digital twin's harvest timing recommendation systematically pushing harvest later at the cost of downstream process failures that get attributed to other causes?
Formula
Notation:
HTAI = ρ(ΔT_harvest, Downstream_Failure_Score)
Where ΔT_harvest = T_actual − T_baseline
Variables:
- ΔT_harvest = harvest timing deviation from historical median (hours)
- T_actual = actual harvest time
- T_baseline = historical median harvest time from pre-twin period
- Downstream_Failure_Score = composite of contamination detection, cell lysis markers, and purification yield loss (each normalized 0–1 and summed)
Worked Example:
28 batches with harvest timing deviation ranging from -2 to +14 hours relative to historical median. Downstream Failure Score computed per batch. Spearman rank correlation: ρ = 0.47, p = 0.012. A moderate positive correlation: later harvest times are associated with worse downstream outcomes. Does not prove causation but is strong enough to warrant investigation.
Edge Cases:
- HTAI near zero = no correlation between timing and downstream outcomes
- Positive HTAI = later harvest correlates with more downstream failures
- Statistical significance (p-value) determines whether the correlation is real or noise
- Minimum 20 batches for meaningful correlation calculation
Sign convention: HTAI near zero is neutral. Positive with statistical significance indicates a problem.
Proposed formula, pending validation pilots.