What It Means to Map Your Genome, and How to Do It Properly
To map your genome today means to sequence your DNA at sufficient depth and align those reads to a reference assembly, producing a per-position record of where you differ from that reference. In practice this is 30x whole-genome sequencing followed by alignment and variant calling. The deliverable you should care about is a set of files: FASTQ for raw reads, CRAM or BAM for aligned reads, and VCF or gVCF for called variants. Anything that gives you a PDF of wellness scores and no underlying files has not mapped your genome in any sense you can work with. The list price for 30x WGS with a full analysis pipeline and data return runs roughly $1,000 to $3,000 as a standalone service, with sequencing-only offerings lower and clinical-grade reporting higher.
The old meaning of mapping, and why the word stuck
Before cheap sequencing, mapping and sequencing were separate problems. Genetic maps ordered markers by recombination frequency in families, measured in centimorgans. Physical maps ordered cloned DNA fragments along chromosomes in base pairs, using restriction sites, YAC and cosmid contigs and later sequence-tagged sites as common landmarks. The Human Genome Project’s early years were largely a mapping project, and the community had to agree on a shared coordinate vocabulary before the sequence could be assembled at all 1. Contemporaneous accounts of the effort framed mapping as the necessary scaffold on which sequencing would later hang 2.
That scaffold now exists as GRCh38 (and, increasingly, the T2T-CHM13 telomere-to-telomere assembly). So when a consumer company says it will “map your genome,” the map already exists and your reads are being placed onto it. The verb survives in bioinformatics as the technical term for read alignment: you map reads to a reference. That is the sense worth caring about, because the quality of that mapping step determines what you can later ask of the data.
The file chain you should insist on
Start with FASTQ, two gzip-compressed files per lane for paired-end reads. A 30x human genome at 2x150 bp typically comes to 90 to 120 GB in total. FASTQ is the only format that lets you re-run everything downstream with a different aligner or a newer reference, which matters because references and callers improve every couple of years.
Alignment produces BAM (binary alignment map) or preferably CRAM. CRAM is reference-compressed and cuts the same data from roughly 60 to 80 GB down to 15 to 25 GB. Keep the CRAM and the exact reference FASTA it was compressed against, because CRAM cannot be decoded without the identical reference bases.
Variant calling produces a VCF. For a single human genome against GRCh38, expect on the order of 4 to 5 million variants, dominated by single-nucleotide variants with several hundred thousand small insertions and deletions. Ask for a gVCF as well: it records confident reference calls alongside variant calls, so you can distinguish “you match the reference here” from “we had no coverage here.” That distinction is the single most common source of false reassurance in personal genomics.
The pipeline we would run
For alignment, use bwa-mem2 against the GRCh38 analysis set with alternate contigs handled correctly, or Illumina DRAGEN if you have access to it. The command is unremarkable:
bwa-mem2 mem -t 32 -R '@RG\tID:s1\tSM:you\tPL:ILLUMINA' \
GRCh38_full_analysis_set_plus_decoy_hla.fa r1.fq.gz r2.fq.gz \
| samtools sort -@ 8 -o you.bam -
Use the decoy-containing analysis set. Do not use a bare chromosome FASTA. Decoy sequences absorb reads from genomic regions absent in the primary assembly and measurably reduce false positives in segmental duplications.
For small-variant calling, we would run DeepVariant over GATK HaplotypeCaller for a single sample. DeepVariant’s convolutional model gives better indel precision in homopolymer runs, and it needs no per-cohort recalibration. Validate against GIAB (Genome in a Bottle) truth sets with hap.py restricted to the high-confidence BED, and expect SNV F1 above 0.995 at 30x. Structural variants need their own callers: Manta plus Smoove on short reads, with the caveat that short-read SV recall is poor, roughly half of what long reads achieve. Repeat expansions require ExpansionHunter with a locus catalog. CYP2D6 and other pharmacogenomic star alleles require specialized callers such as Cyrius, because the gene sits next to a near-identical pseudogene that ordinary aligners cannot resolve.
Annotate with Ensembl VEP, pulling in gnomAD allele frequencies and ClinVar clinical assertions. Add a predictor such as REVEL or AlphaMissense. Then filter aggressively, because the overwhelming majority of your 4 million variants are common and uninformative. The population-scale approach to individual variants is to start from frequency, then consequence, then curated assertion.
Where the map is wrong
Mean coverage of 30x does not mean uniform coverage. Roughly five percent of the genome is inaccessible to short reads at useful depth, including many GC-rich first exons, the centromeric and acrocentric regions, and gene families with recent paralogs such as SMN1/SMN2, PMS2, and the HLA locus. Your VCF will be silent there, and silence reads as normal unless you check the gVCF depth blocks.
Reference bias is the second failure mode. GRCh38 is a mosaic of a small number of donors, and alleles absent from it are harder to detect. Ancestry-matched references and pangenome graph alignment reduce this, and it is worth re-running against T2T-CHM13 if you care about previously unmappable regions.
The third is interpretation rather than alignment. Polygenic risk scores are ancestry-sensitive: a score trained in European cohorts loses predictive power in other populations, which is why groups have built and validated population-specific scores, for example a type 2 diabetes score developed in an Indian cohort 3. If you compute a PRS with plink2 --score, check that your genotype ancestry matches the training GWAS, and report the result as a percentile within a matched reference panel rather than as an absolute risk.
Finally, DNA alone tells you about potential. It does not tell you about activity. Deciding which transcript a gene produces requires spliced alignment of RNA reads to the genome, the problem that tools like GMAP were built to solve 4. It matters because alternative splicing is not a rare curiosity: analysis of expressed sequences found evidence for it in roughly half of human genes 5. A variant near a splice site is often uninterpretable from the genome alone and obvious from your own RNA.
What a mapped genome supports, and what it does not
A well-mapped genome supports carrier status for recessive conditions, pharmacogenomic star-allele calls, and screening of the roughly eighty genes on the ACMG secondary-findings list. It also supports ancestry and relatedness inference as well as polygenic scores with the caveats above. It also becomes the fixed backbone for every later measurement: RNA expression, proteomics, and biomarker trajectories all read differently when you know the underlying genotype.
It does not diagnose anything. A pathogenic or likely pathogenic variant found in research-grade data needs orthogonal confirmation in a CLIA-certified laboratory and interpretation with a clinical geneticist or genetic counselor, who will weigh your family history and phenotype against the variant. We would not act on any single VCF line without that step.
Questions people also ask
What does it mean to map a genome? Historically, to build genetic and physical maps ordering markers and cloned fragments along chromosomes so a sequence could be assembled against a shared coordinate system 1. Currently, to align sequencing reads from one person to an existing reference assembly and record the differences.
How much does it cost to map your genome? Sequencing-only 30x whole-genome services are available for a few hundred dollars. Full pipelines with clinical-grade reporting, confirmatory testing, and data return sit in the low thousands. The consumable cost of sequencing is now a minority of the total price.
What is the most reliable DNA testing company? Reliability is a property of the assay and the deliverable. The brand does not determine it. Ask three questions: is it whole-genome sequencing or a genotyping array, what is the mean and minimum coverage, and do you receive FASTQ, CRAM, and gVCF. Array-based tests genotype roughly 600,000 to 1 million sites and impute the rest, which is adequate for ancestry and common-variant scores and unreliable for rare variants.
Is Mapmygenome legit? It is an established Indian genomics company offering array-based and sequencing panels with counseling. As with any provider, evaluate it on the assay type and the raw data you receive. Also check whether findings are confirmed in an accredited laboratory before anyone acts on them.
What is the purpose of genome mapping? For a population, to build the reference coordinate system that all later work depends on. For an individual, the purpose is to produce a permanent and re-analyzable record of your variants. You can query that record again as reference assemblies, callers and evidence databases improve.
Footnotes
-
Maynard Olson, Leroy Hood, Charles Cantor, et al. A Common Language for Physical Mapping of the Human Genome. Science, 1989. https://doi.org/10.1126/science.2781285 ↩ ↩2
-
P.S Harper. Mapping and sequencing the human genome. Endeavour, 1989. https://doi.org/10.1016/0160-9327(89)90035-5 ↩
-
Sandhya Kiran Pemmasani, Shravya Atmakuri, Anuradha Acharya. Genome-wide polygenic risk score for type 2 diabetes in Indian population. Scientific Reports, 2023. https://doi.org/10.1038/s41598-023-38768-5 ↩
-
T. D. Wu, C. K. Watanabe. GMAP: a genomic mapping and alignment program for mRNA and EST sequences. Bioinformatics, 2005. https://doi.org/10.1093/bioinformatics/bti310 ↩
-
B. Modrek. Genome-wide detection of alternative splicing in expressed sequences of human genes. Nucleic Acids Research, 2001. https://doi.org/10.1093/nar/29.13.2850 ↩
Pothos builds longitudinal molecular profiles of individuals: whole-genome sequencing, single-cell RNA sequencing, and plasma proteomics, delivered as a detailed report, every data file, and stored cells. Build your profile.