Comparison of DNA compression performance on a standard dataset.

Note: Values in each column (except for ) refer to bit per base (bpb). refers to the respective minimum match length in the first pass and ‘Default’ column represents the compression results of COM-NONREF with default parameter (i.e., 25). The ‘’XM500’’ refers to the XM algorithm using at most 500 experts. ‘’COMPACT-seq’’ indicates one mode of our proposed method that uses only traditional sequential models followed by logistic regression model. The value of the best performance model is in bold font.

注：除（原文此处缺失列名）外，各列数值均以每碱基比特数（bit per base, bpb）为单位。该列指代首轮匹配阶段的最小匹配长度，而"Default"列代表采用默认参数（即25）的COM-NONREF算法的压缩结果。"XM500"指代最多使用500个专家的XM算法。"COMPACT-seq"为本研究提出方法的一种变体，仅采用传统序列模型后接逻辑回归（logistic regression）模型。性能最优模型的数值以粗体标注。