Do Read Errors Matter for DNA Assembly?
summary
Assumes a dense read model (a read starting at every base, cyclic) and adversarial erasures (≤ D erasures per read and ≤ D erasures per base, i.e. erasure rate D/L). Proves a combinatorial spectrum error correction guarantee: from those erased reads one can recover an error-free set of (K+1)-mers when L exceeds the original critical read length plus a term depending on the sequence’s count of approximate repeats; empirical tests on real genomes show the required L grows modestly (roughly up to ~50% in worst examples).
AI-written summary · about