CycSim

Long-read sequencing data contain context-dependent errors, where certain bases are more likely to be misread depending on their surrounding sequence. Most existing simulators introduce errors randomly, which overlooks these error biases and only approximates the overall error rate. CycSim takes a different approach by modeling errors in a k-mer–dependent manner, enabling more realistic and biologically accurate error simulation. CycSim is easy to train and supports all types of long-read sequencing data. It currently provides pre-trained models for BGI CycloneSEQ, PacBio HiFi, and Oxford Nanopore Q20 data. Users can also quickly train their own custom models using a BAM file of reads aligned to a reference genome.