Fig. 9
From: Magic-BLAST, an accurate RNA-seq aligner for long and short reads
Number of matches (X-axis) versus mismatches (Y-Axis) for iRefSeq and seven experimental sets. The aligner with the highest number of matching bases is most to the right and is the most sensitive. Preferably, it should be more to the top, as mis-mappings will increase the mismatch rate and pull the aligner downwards. For each set, the number of mismatches per kilobase for the best aligner is shown in the color of the aligner. It likely reflects the actual mismatch rate in the sample. The iRefSeq sequences match the genome exactly, so the truth (dark blue square) is in the top right corner. The Magic-BLAST red square overlaps the truth. HISAT2 relaxed (grey) is the second most sensitive aligner (second most to the right) but has over 60,000 mismatches. STAR long has less mismatches but is by far the least sensitive. In all runs with long reads (3 PacBio, Roche and 2 long Illumina), Magic-BLAST has the most matching bases yet keeps a controlled mismatch rates. In Illumina 101 + 101, STAR long has slightly more matching bases but with a much higher mismatch rate. Notice that the PacBio brain and testes runs had been edited by the submitters using Illumina runs, which explains their low mismatch rate