Producing a biologic (an antibody, an enzyme, a vaccine antigen) starts by transfecting a host β typically CHO or HEK cells β with the gene of interest, then screening a large pool of resulting clones for the handful that make the most product per cell per day. Each well starts with a titer drawn from a log-normal distribution, the standard empirical shape for clonal expression variance:
T_i = exp(ΞΌ + ΟΒ·Z_i), Z_i ~ N(0,1)
Screening keeps only the top fraction by stringency β the same rank-and-cut logic as real high-throughput titer assays (ELISA, Octet, imaging-based picking). The ranked bar panel makes that cutoff visible directly instead of only inferring it from a 3D plate.
The selected clones are then expanded across passages. Two independent processes erode productivity over time, and this simulator models both:
- Genetic drift β small multiplicative noise applied to titer every passage, T(t+1) = T(t)Β·(1 + driftΒ·Z), reflecting gradual copy-number and expression-cassette variation.
- Epigenetic silencing β each passage, every clone has an independent probability of promoter methylation collapsing its output toward near-zero, T(t+1) β 0.05Β·T(t). This is why real cell-line development runs extended (often 8β12 week) stability studies before locking a production clone.
The plate heatmap encodes titer as well size and colour (large/green = high producer, small/red-grey = weak, not-selected or silenced). Hover a well to read its exact titer. The history chart plots mean titer of the surviving population and the count of "stable producers" β clones still above half their post-screening titer, the practical bar a manufacturing clone must clear β across every passage advanced so far.