Abstract
<title>Abstract</title> <p>SCRaMbLE generates structural diversity in synthetic yeast genomes, but endpoint sequencing mainly captures viable clones and leaves trajectories removed by viability selection largely unobserved. To explore this missing landscape, we developed a risk-aware lox-order stochastic framework for model-defined viability loss and survival–diversity trade-offs during intrachromosomal SCRaMbLE. Each chromosome was represented as a dynamic sequence of lox-bounded segments, with lox order, segment orientation, junctions, and ORF copy numbers updated after each deletion, inversion, or tandem duplication. An Essentiality Gate classified a trajectory as gate-failing when any retained essential ORF reached zero copies. We evaluated two lox-pair sampling schemes, four distance-decay parameters, and 11 event probabilities across 88 scenarios, each comprising 10,000 trajectories and up to 20 event opportunities. Increasing event probability consistently reduced the Gate-passing fraction while increasing copy-label-invariant structural diversity among passing endpoints. All 11 Pareto-optimal settings arose under linear-distance sampling at α = 2.0. At a prespecified Gate-passing threshold of 0.70, the most diverse admissible setting occurred at \(\:{p}_{\text{e}\text{v}\text{e}\text{n}\text{t}}=0.10\), with a passing fraction of 0.7285 and rarefied structural effective Shannon diversity of 456.16. Seed and Gate-definition analyses preserved this trade-off. Under the zero-copy rule, Gate failures were necessarily deletion-triggered and concentrated in recurrent essential-gene-containing intervals on SynII. Exact state replay showed that 44.73% of Gate-failing trajectories contained at least one preceding inversion or duplication; prospective all-pair analysis further showed that inversions increased one-step Gate-failing deletion hazard in 41.66% of sampled states, whereas duplications decreased it in 96.67%. Partial Hi-C information produced the highest pooled Pearson correlation (\(\:r=0.4509\)), but did not significantly outperform matched distance-based sampling (\(\:p=0.2865\)); hotspot ranking remained near random and chromosome-wise holdout correlations were weak. The framework provides a reproducible basis for comparing rearrangement regimes, resolving state-dependent cascades, and generating interval-level hypotheses for experimental calibration.</p>