This question has also been asked on Biology SE

When a raw number like 1.3 percent is stated, it is confusing without context,and can be misinterpreted. So does 1.3 percent mean that 1.3 percent of the letters/nucleotides/bases of my genome consists of established (likely/high confidence) Neanderthal markers or SNPS? Does it mean 1.3 percent of my entire genome consists of these Neanderthal markers, i.e 1.3 percent of 3.1 billion or 6.2 billion base pairs, or is it 1.3 percent of protein coding regions, or exons?**

How do they obtain these markers for Neanderthal ancestry to begin with? I know the Neanderthal genome has been sequenced from bone fragments. Do they compare SNPs and variable regions in the genomes between Neanderthal and Modern Human population genomes to see the relative frequencies of these variations/SNPs. Then likely Neanderthal markers are flagged by statistical models and algorithms..for example combinations of SNPs that appear infrequently or hardly at all in the reference Sapien population genome but occur frequently in the Neanderthal reference genome..that marks these SNPs/variable regions as of likely Neanderthal origin? Could you please describe some of these statistical methods and algorithms? Can anything else explain the "markers", or SNPs other than interbreeding with Neanderthal (such as mutations)? Please illustrate using some hypothetical alignments like AAGGGT..vs GAGGGT..Please give a clear, concise and authoritative answer of high confidence. It is okay to be detailed and technical, although building up to complexity so that it can be understood with relatively little familiarity with population genetics and Mathematics.


3 Answers 3


I ran a basic search of 23andme and it looks like they use known variant loci from the Neanderthal genome. Also, they expect a 2-2.5% overlap. See:

  1. https://customercare.23andme.com/hc/en-us/articles/212873707-Neanderthal-Ancestry-Report-Basics
  2. https://blog.23andme.com/articles/new-neanderthal-report (Section "The Science behind the report")
  3. https://permalinks.23andme.com/pdf/23-05_neanderthal_ancestry_inference.pdf (Eric Durand's white paper on the technical details)
  • $\begingroup$ So the percentage they give corresponds to the percentage of the specific set of variant loci that a test sample has in common with a Neanderthal reference genome? What does the actual percentage mean, exactly? $\endgroup$
    – terdon
    Apr 8 at 12:40
  • $\begingroup$ @terdon I believe the percentage gives the fraction of your genome derived from a Neanderthal genome. $\endgroup$
    – user172818
    Apr 8 at 15:58
  • $\begingroup$ @user172818 yeah, but fraction of what, exactly? Percent of similarity between coding regions? Between known transcript regions? Genes? Across the whole genome including non coding and intergenic areas? That is what the OP is asking, as far as I can tell, and the answer isn't clear (to me) so far. Perhaps the answers are in the given links, but if so, they are not here. $\endgroup$
    – terdon
    Apr 8 at 16:01
  • $\begingroup$ @terdon Whole genome $\endgroup$
    – user172818
    Apr 8 at 16:12
  • $\begingroup$ Do you mean the results from the 'whole-genome' SNP (genotyping) chips? $\endgroup$ Apr 9 at 5:08

Based on reading a similar post it seems like if the report says I am 1.3 percent Neanderthal then it does NOT mean 1.3% of my base pairs or SNPs of my genome (or exons or protein coding region or total number of my SNPs) has a "Neanderthal" label inferred, rather some kind of statistical inference is made as a whole.

The author suggests when it comes to 23AndMe that my genome is segmented into chunks of a chromosome (containing a set of SNPs, 23AndMe does genotyping only reads parts of chromosome where bases are known to vary.)

Technical details could be referred to the following sites: https://academic.oup.com/bib/article/19/5/765/3038466



Since 23andMe is using v5 genotyping chip, the fraction of Neanderthal variant could be calculated. Moreover, the basis of v5 genotyping chip is related to the concept that DNA-based polymorphisms (SNPs or SNVs) which differ substantially in frequency among human populations can be employed to infer ancestry. With 23andMe, an average of 236 Neanderthal SNPs/SNVs could be obtained with the analysis based on a selected group of consented customers.

And these chunks are fed into a Machine learning classification algorithm like Support vector machine (SVM). The Machine learning algorithm was trained chunks of genome of reference population sets with chunks of known ancestry (like "Scandinavian" or "South European" or "Sub Saharan African").

Then the classification algorithm makes a multilabel prediction... lets say the algorithm predicts that a particular segment is 95 percent likely to be human and 5 percent Neanderthal..and maybe the next segment has a 2% probability of being human and 98 percent Neanderthal..Then the algorithm looks for segments of high confidence (>90 percent or 70 percent whatever threshold you set) and then add them up.

For example if in a toy example genome has 9 segments of the same size has a greater than 90 percent probability of being estimated "human", and 1 segment with greater than 90 percent of being estimated Neanderthal by the classification algorithm than I would be considered to be 10 percent Neanderthal...

The author suggests this is the method used to infer ancestry like "Norwegian" and "South European" and "Sub Saharan African" by 23And Me..but will it be similar for inferring Neanderthal ancestry?

Calculating ancestry percentage is not seem as simple as counting up number of "Neanderthal" SNPs/Markers and dividing by your total SNPs or base pairs in your genome/exon...Rather an inference of ancestry mix needs to be made based on reading SNPs on segments of chromosomes and comparing it to reference data sets via classification algorithms.

  • 1
    $\begingroup$ Also could someone clarify what "segments" or chunks mean. when someone mentions "..take the genome of target individuals and split it up into chunks along the genome". I know 23 and Me does genotyping not WGS or whole exome sequencing..so these segments or chunks must consist of a a set of SNPs..presumably, rather than a continuous segment of the Genome including invariable/conserved region...? $\endgroup$ Apr 8 at 23:28
  • 1
    $\begingroup$ You are overthinking it. What method they use doesn't matter much. The goal of ancestry estimate is to get genome fraction. $\endgroup$
    – user172818
    Apr 9 at 1:06
  • $\begingroup$ 'The most commonly used criterion for SNP (genotyping) chip evaluation is global coverage, defined as the fraction of common SNPs that are tagged by the SNPs on the chip.' Do you mean the fraction of common SNPs is the same as the number of SNPs tagged by SNVs on the SNP (genotyping) array? $\endgroup$ Apr 9 at 4:43
  • $\begingroup$ @SanjayBiswas You are right. Current 23andme is using their v5 genotyping array and genotyping arrays could measure up to millions of SNPs across the genome as illustrated in this table --> nature.com/articles/s41431-021-00917-7/tables/1 $\endgroup$ Apr 9 at 4:55
  • $\begingroup$ @SanjayBiswas Please also refer to this slide shown in the powerpoint of Lynn Jorde --> youtu.be/PzG99bexziA?t=3712 $\endgroup$ Apr 15 at 5:09

How do they obtain these markers for Neanderthal ancestry to begin with? I know the Neanderthal genome has been sequenced from bone fragments

  1. Since ancient nuclear DNA is scarce, enough preserved molecules are needed.
  2. The level of scarcity about ancient nuclear DNA : often, 1% or fewer of the total DNA fragments from an extraction among soil microbes and other sources
  3. After extraction and sequencing, distinguish the genuine DNA fragments from those of modern contaminants and from those of microorganisms that are co-extracted from the sample
  4. Since ancient DNA has characteristic damage patterns, e.g. a high degree of cytosine deaminations on the five prime ends, and at the three prime end, there is a high frequency of Gs changing to As.
  5. Match these ancient DNA fragments to a map of the complete human genome reference sequence; this allowed the book of the target's genome to be assembled from the millions of random sequence fragments recovered from the bones
  6. The more times we had a read confirmed by multiple fragments, the more confident the sequencing could be in distinguishing the true DNA sequence, it is the sequencing depth issue
  7. Then genotype samples to discover, filter, and analyze variants.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.