Alignment validation and performance

July 16, 2026 · View on GitHub

rammap 1.0.0 vs minimap2 v2.31

System


CPU	Intel Xeon Gold 6140 (2.3 GHz, AVX512)
RAM	32 GB allocated
Threads	8

Publicly available ONT, PacBio (HiFi and Onso), Illumina (PCR-free and Hi-C), BGISEQ, and Element AVITI DNA sequencing data and ONT RNA-sequencing data were evaluated across various presets and alignment modes. Datasets are drawn from the Genome in a Bottle (GIAB) reference releases and the Human Pangenome Reference Consortium (HPRC) production data releases, supplemented by a public direct-RNA reference run for ONT's RNA004 chemistry.

Dataset	Source	Access
ONT R10 (sup)	HG002; GIAB	`s3://ont-open-data/giab_2025.01/basecalling/sup/HG002/`
ONT ultra-long	HG002; GIAB	HG002_ONT-UL_GIAB_20200204.fastq.gz
PacBio HiFi (Revio, May 2024)	HG002	hg002v1.0.1_hifi_revio_pbmay24.bam
PacBio Onso	HG002; HPRC	`s3://human-pangenomics/T2T/scratch/HG002/sequencing/onso/Broad-Onso-HG002/HG002/`
Illumina PCR-free 2×250 bp	HG002; GIAB	HG002.GRCh38.2x250.bam
Illumina Hi-C	HG002; HPRC	`s3://human-pangenomics/working/HPRC_PLUS/HG002/raw_data/hic/downsampled/`
Element AVITI 2×150 bp, ~500–600 bp insert	HG002; HPRC; AVITI chemistry	`s3://human-pangenomics/T2T/scratch/HG002/sequencing/element/trio/HG002/ins500_600/`
Element AVITI 2×150 bp, ~1 kbp insert	HG002; HPRC; AVITI chemistry	`s3://human-pangenomics/T2T/scratch/HG002/sequencing/element/trio/HG002/ins1000/`
BGISEQ-500 PCR-free 2×150 bp (L1)	HG002; GIAB	BGISEQ500_PCRfree_NA24385_CL100076190_L01_read_{1,2}.fq.gz
ONT direct RNA (RNA004, sup)	Universal Human Reference (UHR) RNA	PNXRXX240011_dorado7213sup.fastq.gz
GRCh38 (human reference)	Genome Reference Consortium	GCA_000001405.15_GRCh38_genomic.fna.gz
GRCm38 (mouse reference)	Genome Reference Consortium	GCF_000001635.20_GRCm38_genomic.fna.gz
T2T-CHM13v2.0 (human reference)	Telomere-to-Telomere Consortium	GCA_009914755.4_T2T-CHM13v2.0_genomic.fna.gz

Performance Comparison (8 Threads)

CPU time, wall time, and peak RSS for rammap (rm) vs minimap2 (mm2). Bold marks the better outcome per metric (lower is better). In all cases, rammap produces identical alignment output to minimap2 (with the exception of SAM headers including the tool name itself).

Long-Read Presets

Dataset	Preset	rm CPU (s)	mm2 CPU (s)	rm Wall	mm2 Wall	rm Mem	mm2 Mem
ONT R10	`-cx lr:hq`	230538	260983	8:16:02	10:23:27	13.0 GB	16.4 GB
ONT UL	`-cx map-ont`	664349	684362	25:30:04	26:00:17	16.6 GB	14.1 GB
ONT UL (1M reads)	`-cx ava-ont`	30446	31645	1:10:37	1:16:12	47.2 GB	38.1 GB
PacBio HiFi	`-cx map-hifi`	274260	318231	9:53:33	11:08:45	12.9 GB	16.1 GB

Splice / RNA Presets

Dataset	Preset	rm CPU (s)	mm2 CPU (s)	rm Wall	mm2 Wall	rm Mem	mm2 Mem
ONT RNA	`-cx splice -uf -k14`	121612	131063	4:23:17	4:37:06	21.9 GB	22.5 GB

Short-Read Presets

Dataset	Preset	rm CPU (s)	mm2 CPU (s)	rm Wall	mm2 Wall	rm Mem	mm2 Mem
BGI PCR-free	`-ax sr`	432217	307897	15:16:59	12:33:02	14.9 GB	12.6 GB
Illumina PCR-free	`-ax sr`	355364	306620	12:44:05	11:02:44	15.0 GB	12.9 GB
PacBio Onso	`-ax sr`	161168	152442	5:44:25	5:20:48	14.9 GB	12.5 GB
AVITI (1 kbp insert)	`-ax sr`	620332	430574	21:45:35	17:12:46	14.9 GB	12.6 GB
AVITI (500–600 bp insert)	`-ax sr`	760693	608294	27:21:27	27:05:20	14.9 GB	12.5 GB
Illumina Hi-C	`-ax sr --frag=no`	2410520	1727916	83:41:55	64:32:11	15.6 GB	13.0 GB

Assembly Presets

Dataset	Preset	rm CPU (s)	mm2 CPU (s)	rm Wall	mm2 Wall	rm Mem	mm2 Mem
GRCh38 / GRCm38	`-x asm20`	1522	2004	0:08:13	0:10:15	23.5 GB	21.0 GB
GRCh38 / T2T-CHM13	`-x asm5`	6894	7084	0:54:06	1:00:54	30.3 GB	28.2 GB