23andMe’s New Custom Chip

23andMe will soon be genotyping customers with a newly designed custom microarray chip.23andMe_Logo_blog

The new version – along with completion of additional lab space – will ultimately improve both the efficiency and speed by which we will be able to process samples and then return genetic data to our customers.

Unlike previous changes to our chip, this new version, dubbed v4, consists of a fully custom panel of probes for detecting single nucleotide polymorphisms (SNPs) selected by our researchers. The selection was made to maximize the number of actionable health and ancestry features available to customers as well as offer flexibility for future research.

Unlike previous changes to our chips, this one will not significantly alter what is reported to customers, so there would be little reason to retest on the latest platform.Among other things, the new chip includes both common variants that provide genome-wide coverage and rare variants that are found in certain medically relevant genes and exons. Researchers also ensured that the chip covers SNPs with known disease associations as well as those variants that are known to be involved with drug metabolism and side effects.

In addition, v4 includes SNPs that our researchers use to assign genetic ancestry using autosomal DNA markers as well as variants in both mitochondrial DNA and in the Y chromosome. The custom chip will also allow our researchers to better impute many variants present on larger chips. Imputation is a powerful statistical method that allows researchers to fill in missing data points.

Researchers use publicly available full-sequence data, 23andMe’s own data and other sources to impute the genotype at specific locations.While the new chip was designed to be compatible with all our current features, there may be some small differences in some reports. 23andMe will inform our customers of those differences in community posts.If you haven’t seen it already, the video below fills you in on how a sample is processed at the lab.

  • rasheemo

    when will this new chip go into use?

    • 23blog

      We will likely start using the chip in the next few weeks.

      • Alpentrudel Alpi

        How many SNP are on this V4-23anmde-chip?

        • 23blog

          There are more than 600,000 SNPs on V4.

          • Mike

            “More than 600,000 SNPs” could be quite a bit less than the 967,000 SNPs the V3 chip has. Sounds like some things could be left out. If so,

          • 23blog

            Mike, We designed the chip so that we’d be able to report out the same information to customers. The SNPs that are covered also give our scientists the ability to impute data that you would need a much larger chip to cover. There are SNPs that are on v3 that are not on v4 but we that should not substantially change the reports customers receive.

  • Nathan M

    I’m assuming it’s backward-compatible with v3. How many SNPs will the new chip genotype? And will outstanding kits not yet returned be run on the new ship, or is there a stock of v3 chips to deplete yet?

    • 23blog

      Yes, it is backward compatible with V3. The new chip has more than 600,000 SNPs. In answer to your last question it sort of depends on when the sample is sent in. We are transitioning to the new chip but are still using v3.

      • Nathan M

        Is that 600,000 total or additional SNPs? By my count, v3 has more than 960,000 SNPs.

      • R Stewart

        V3 offers 967,000 SNPs. Does this mean that V4 will be looking at a lot less SNPs? More than 600,000 does not sound anywhere near the almost 1 million tested with V3.

        • 23blog

          The new chip is looking at fewer SNPs but because it is a custom chip, we are able to glean the information we need to report out the same information as the previous version. We can do this and improve the efficiency of processing the data.

          • R Stewart

            It’s a huge downgrade so please don’t paint it as an improvement. Our family will be showing our displeasure at your business decision by testing at other companies.

          • R Stewart

            Be honest. Admit it is a cost saving measure and not an improvement. It is a serious downgrade and 23andme should be ashamed of itself for doing it. Stop offering the discounts, raise prices. It was unwise to go with such a low price point when you know it was not sustainable.

  • Dustin Haley

    if our specimen has not been analyzed yet, can we hold off to have it use the new chip?

    • 23blog

      If you’ve already sent in your sample, you cannot put the processing on hold.

  • Bri Bri

    I would love to see a list (I understand it would be very long!) of the actual SNPs being tested by the new chip, or at least those being tested that are not reported on with the V3 chip. That might seem anal, but I don’t have to break too many piggie banks to get another $99 together. There might even be just one SNP that V4 covers that V3 does not that is worth $99! (People are still paying hundreds JUST to find out their APOE status, for ex.) Thanks, GB

  • 23blog

    No it is not a downgrade. It does allow us more efficiently process the samples. The difference between v3 and v4 is that our researchers picked all the SNPs on the chip. It is totally custom. That means they were able to select the most actionable variants. That allows 23andMe to continue to report out to customers what they get from v3, but it allows means that the variants we’re looking at are also the most useful for leveraging additional data. With this chip we can impute – filling in missing data – much more than we could with v3.

    • Jj beeper

      Hmmm…. Is it really a “totally custom” chip? From my understanding, the majority of SNPs on the Illumina chip is a fixed/standard selection of SNPs, and 23&Me can then add some (but a minority) of custom SNPs of interest. Perhaps I am mistaken. Could you clarify? Thank you.

      • 23blog

        This is an all custom chip.

    • randommomster

      Please explain how you are going to “report out to” me on some 400,000 SNPs that you’re not going to be testing? What additional data (NOT imputed information) can you get from looking at only 2/3s of the data points? I am really not getting how this is an improvement for anything other than your bottom line. What SNPs are gained on v4, and how do they make up for such a stunning loss of data?

    • randommomster

      It sure seems like a downgrade to me. Still waiting for an answer on the following, from nearly a week ago:

      Please explain how you are going to “report out to” me on some 400,000
      SNPs that you’re not going to be testing? What additional data (NOT
      imputed information) can you get from looking at only 2/3s of the data
      points? I am really not getting how this is an improvement for anything
      other than your bottom line. What SNPs are gained on v4, and how do
      they make up for such a stunning loss of data?

  • 23blog

    Bri Bri, The difference between v2 and v3 is that with that change in the chip we did indeed report out information new information (APOE, being the best example). The new chip is meant to be backward compatible so that customers will get the same reports back that they get with v3. There are some different SNPs in there and it may be that some of those are important for individuals, but at the moment we don’t think it is necessary for v3 customers to upgrade.

    • Bri Bri

      Thanks for the useful info! I understand it probably wouldn’t be necessary for V3 customers to upgrade.

      But I’m not sure about the answer to my question: will a list of the different SNPs be made available? (I suppose it would be easy enough to ask a V4 customer to post it somewhere, of course!)


  • Massimo Concas

    I think it would be useful to better highlight those ~400 000 SNPs left behind so we can understand which areas aren’t covered or worsely covered now. Because obviously with so many less SNPs something has to give so to speek.

    P.S.: I’m kinda glad that I will most likely be tested with the V3

    • 23blog

      This is a 100 percent custom chip design. Our entire genotyping array is now all hand-curated by 23andMe scientists. This has eliminated some of the SNP locations previously captured that were not informative to our reports or research and has enabled us to maximize the number of actionable health reports we can provide to our customers.

      Our chip now probes for about 600,000 to 700,000 locations for each customer – capturing a more refined set of SNPs than our previous chip.

      In addition, changing formats enables us to be more efficient. We now have the ability to process 24 samples per chip instead of 8 samples per chip on v3. This change significantly increases our lab capacity and will help increase the turn around time of results to customers.

      • Massimo Concas

        Thanks for the answer.
        How far, or rather close, do you think you are to providing the option for a full sequencing at an affordable price?

        • 23blog

          Massimo, Thanks for the note. We don’t have an immediate plan. However we are preparing for sequencing and have done small sequencing projects. I wish I could give you a better answer, but at the moment all I can say is we are preparing for that time.

  • IM

    Hey 23Blog,

    I understand that this is a 100% custom chip design, and that it can improve turn around time dramatically. Thank you for the effort.

    I would really be interested in knowing what some of the SNPs being left out include (like you mentioned: “There are some different SNPs in there and it may be that some of those are important for individuals”). I understand that you may not be able to list each one, but if you were to select the top 5 that may affect the most number of customers which would they be?

    Also, are there any new SNPs being looked at? Or are the ~600K covered by v4 included within the ~960K of the v3?

    Thank you!

    • 23blog

      Thanks for the note. There are some SNPs that we are looking at on v4 that are not on v3. You can see the types of reports that you would receive from your data, which are the same as the reports that v3 customers received. Customers can also download their raw data to look at specific SNPs.

    • Carl

      I’m not from 23andMe, so I can give you a more useful answer. Some of the SNPs being left out were “no call” SNPs that often gave no result or were unreliable. I don’t know about the other ones, but because 23andMe completely designed this for their needs, I’m guessing they are SNPs that are less likely to provide health information, ones that are less useful for ancestry, and ones that tell you the same information as other SNPs they are already measuring.
      Yes, there are a lot of new SNPs being looked at. Especially on the Y chromosome and the mitochondria, which both have more SNPs than they did before. They are trying to be able to detect all the precise haplogroups accurately, which will allow them to reasonably predict (aka “impute”) the other Y chromosome or mitochondrial SNPs they don’t measure. The new chip will be more compatible with mitosearch.org.
      The autosomal DNA will have new SNPs too (but less of the old ones and less overall), which are also designed to make it easier for them to predict other SNPs that aren’t being measured. In other words, they are trying to cut down on redundant SNPs and only use predictive ones.
      It’s unlikely to affect the health and trait information they currently provide, since that only uses a very small number of the SNPs they measure, and I’m sure they want to keep providing that information (if the FDA allows it).

  • Timothy Woods

    Are some imputed SNPs used to produce the results reported for the v3
    chip today, or is this a new practice starting with the v4 chip?

    • 23blog

      Yes we do some imputation currently with v3.

  • Craig Hillemann

    The current V3 chip is generally known to be a SNP superset of chips used by AncestryDNA and FTDNA, and therefore raw data from all three can be compared at GEDMatch.com. Will raw data from the V4 chip be equally useful for this purpose?

  • Sonia L. Walker

    1) You mentioned v3 compatibility, but is it compatible with v2, or will v2 and v4 kits miss potential relationships?

    2) Please specify how this will negatively impact on those of us interested in genealogy.

    2) Since you’ve arrived at your optimal customized SNP search, I assume there’s no need to do any more health surveys. You have all the answers now.

  • villandra

    I’m unclear on whether the new test tests less, tests more, or what, in relation to v. 3. Also I really want to know how it exactly compares to v.3, especially when it comes to Y DNA and mitochondrial DNA. For instance there are a host of new mitochondrial SNPs that define many new subclades, such as H1am1; will those be included? What about the new Y DNA SNPs? I was shocked to find that for a R1b1a2 cousin, the degree of resolution is something like P312, which includes much of central and northwestern Europe, when there are a multitude of subclades known.

    For the medical SNPs, PLEASE don’t tell me they are going to home in on the SNPs in their reports or what 23andMe researchers are focusing on or whatever, and there is no longer the wealth of further health related SNPs in the raw data for those who care to do in depth research.

  • villandra

    OK, I’ve now read the discussion.

    I’d buy the test if it tested MORE SNPs, not less.

    • John Doe

      500,000+ SNPs tested on the v4, almost 1,000,000 on the older chips.

  • villandra

    Currently users upload their 23andMe results to Gedmatch for comparison to the highly similar Family Tree DNA and Ancestry results; I guess this would make that impossible. Don’t look to sell a whole lot of v.4.

  • 23blog

    We are designing our product to improve results for our customers, not for compatibility with other services.

    • Julia

      I decided to use your service just because the results can be transported to other services. I will be very disappointed if I can not do that 🙁
      My test I just recently sent – it will be tested with the new chip?

      • Nick

        Julia I agree. 23andMe was at the top of my list for DNA testing until I read what appears to be a lack of data portability.

    • IM

      Compatibility with other services is beneficial for your customers.

    • randommomster

      Compatibility with other services IS an “improvement”, a feature if you will. That capability is why I chose 23and me, and other of your blog articles tout the larger number of SNPs as a competitive advantage! So far, I am not seeing any advantage to _me_ in v4, only to you. Please reconsider, or at least consider ways in which those of us primarily interested in the ancestry information can be tested with v3.

    • The Other Karen

      For the genetic genealogists among us, being able to search for relatives among those who did NOT use 23andMe, IS a benefit, and makes us see 23andMe in a good light, and would cause us to recommend your service to others.

      You might even consider allowing folks who used other services to upload THEIR data to 23andMe, like Family Tree DNA does. It only benefits THEM. That’s why they allow it.

  • Craig Hillemann

    If the V4 chip lacks V3 SNPs that are on plates used by AncestryDNA and FTDNA, and thus the V4 raw data is less useful or cannot be used at all for transfer to GEDMatch.com or FTDNA, will it be possible to continue to have samples tested on the V3 chip at 23andMe, on special request and perhaps modest additional cost?

  • charles000

    I may be the lone voice here of the moment, but I’m less concerned about Gedmatch and ancestry related issues, but am quite interested in the “actionable health” aspects of the 23andMe V4 upgrade. Quasi recreational reading for me is drifting through publication sites like GenomeWeb, Nature, etc., articles and papers from which indicate an ever accelerating plethora of genetically definable attributes and characteristics being discovered and researched. Any thoughts you might want to sprinkle in here along these lines?

  • Mike Miller

    I look forward to the V5 chip in which 23andMe will correct the mistake they’ve made here.

    I’m an MBA, I get the efficiency thing. I also get that with the V4 chip there is a lot less value to me and other customers that care about that.

    Efficiency needed to be found in other places. Cutting down on the number of SNP’s was the wrong place to do it and is going to provide your competition with an obvious and easy to explain differentiation point. This will in turn likely allow them to charge more. I know I’d pay extra for more coverage.

  • 23blog

    As we said the v4 chip was not designed to change the results that customers get. Indeed customers should get the same reports that from v4 that they got with v3. That said the chip is totally custom. All the SNPs were picked by our scientists and they made that selection not just to ensure that we could maintain that continuity between v3 and v4, but to give them the ability to impute more information. In addition they picked some SNPs that may be helpful for reporting out information in the future.

    • randommomster

      Imputing information is now made fairly useless for new customers, thanks to your decision to stop providing health reports, right? If all I can get is raw data, how does this “more information” benefit me if you can’t report it?

      • Scott23H

        You are correct, that if you ordered after Nov. 22nd you will not get health reports. However you will get ancestry results and have access to your uninterpreted raw data. If this doesn’t work for you and you ordered on November 22nd or later you can get a refund.

        • randommomster

          Still waiting for a clear answer as to how 600K SNPs is somehow better for me than 900K. It is crucial that those of us who will be unable to get V4’s snazzy new reports get the 50%-more-SNPs provided by the V3 chip.

  • Nick

    Unless we hear differently from you, we’ll assume that the V4 dataset portion of the 23andMe product is NOT COMPATIBLE and NOT PORTABLE to/with GEDCOM or other dna genealogy sites/databases/services/etc. You indicate that the reports are similar to V3 chips but it also sounds like the data output is not compatible. BTW … that is a game-changer for many who valued the genealogy portion of 23andMe and the “hunt” it enabled. BTW … I look forward to being wrong in this observation.

  • randommomster

    Awesome. Not. I JUST dropped mine off at the post office this morning. Will I get the v3 chip? Please? I’m more interested in genealogy and information than I am in making your processes “more efficient” by paying the same amount for 33%-ish less data.

  • AGR

    If it sounds like genetic genealogists are less than excited by the introduction of the v4 chip, you would be absolutely correct. Yet again, it seems that 23 is not listening to important feedback. Leaders of the gg community have posted analysis of the new chip and how it stacks up for this application. (see below) That, combined with the awkward internal communication system for matches and current tools available only from 23, should be enough to make new buyers think long and hard about a purchase from 23. Especially with the hold imposed by the FDA, this does not seem like a move that was made with customers and their applications in mind.

    v4 analysis

    Current communication tool

    There are many other forums, lists, bloggers, etc… talking about all these changes and the impact they could have on use for genealogical purposes.

    Compatibility is only important if you are looking to broaden your pool of potential informative matches, gain every clue you can possibly find to your heritage and ancestry, and access additional very helpful tools.

  • Valentijn

    Will pathogenic missense mutations still be concealed using “i” numbers instead of the standard rs numbers?

    • Scott23H

      We will continue to return raw data to customers.

    • Scott23H

      We did not conceal variants. The reason those variants were identified with “i” numbers is because they were not identified in the various “rs” libraries. We created those numbers as a way to identify them. It didn’t have to do with pathogenic missense.

      • Valentijn

        Yes, some “i” numbers lack rs numbers. But a great many are pathogenic missense mutations with very well known rs numbers. Virtually every known pathogenic missense mutations has been given an “i” number by 23andMe, despite almost all of them having rs numbers elsewhere. Which really makes it look like useful information is deliberately being concealed from casual users.

        • Scott23H

          Nothing is being concealed. And again the i numbers were used because there were not rs numbers, at least at the time.

        • Carl

          I’m just a customer, but:
          rs numbers are assigned by dbSNP. dbSNP only assigns an rs number after the SNP has been proven to vary between people and proven to be measuring what they think they are measuring. When 23andMe adds a new SNP it doesn’t have an rs number yet, so they give it an i number while they wait for dbSNP. After dbSNP has given it an rs number, 23andMe doesn’t immediately rename it because that might break existing software that imports the raw data. I think they probably wait until they are updating their raw data file format and then rename the SNPs to the new rs number. That’s my guess.

  • Scott23H

    The new chip will still allows us to provide the same health reports as V3. Obviously with the FDA decision individuals who purchased after Nov. 22nd will not get health reports. To be honest I don’t know how many individuals who purchased before Nov. 22nd will be on V4.

    • randommomster

      I am more concerned about those of us who purchased after, since we don’t get whatever benefits there may be in the new reports. You’re asking us to make do with only 2/3 the raw data. In fact, I would ask you to reserve the remaining V3s for the post-November 22 crowd for that very reason.

  • Scott23H

    I did answer you question, but I’ll say it again. The new chip was designed to allow us to report back to customers all the same information we reported using v3. It was designed to allow for better imputation. Imputation would allow us to report additional information in the future. That said the new chip is coming on line just as the FDA has ordered us to stop reporting health results to customers, so customers who purchased after November 22nd will not be getting health reports.

    • randommomster


      I wish you had answered me, but you merely replied.

      Your reply discussed reports that can be generated through _imputing_.

      That wasn’t my question.

      I am asking about data.

      Two very different things.

      Post-Nov22 people get _nothing_ from imputed health and trait reports, because we don’t get _any_health and trait reports.

      You are trying to tell us that providing us with significantly less data is somehow better. The “improvement” you cite – which might be an improvement for the pre-Nov22 crowd – does not exist for us. We need the 50% more data of V3.

      • Scott23H

        I understand that it’s not the answer you want but I’m answering your question. The chip was designed so that we could give customers the same ancestry and health reports that they get with v3. Yes there are fewer SNPs, but there is not a significant difference in the results. That said the changes with the FDA obviously make that a mute point.

        • randommomster

          Getting closer.

          We’re not talking about a small reduction in SNPs. We’re talking a significant reduction, from nearly a million down to only 600,000. That’s a HUGE cut.

          Many people are concerned about data portability, probably post-Nov 22nders more than others. Some people in this thread have expressed concern that your new data (NOT interpolated) results are no longer standard, and may not be able to be analyzed well by third-party tools. Those concerns were met with “we didn’t design the chip to be compatible with other services” – as though that were a perfectly understandable and good thing.

          You guys are painting this change as a win. It MAY have been a win for more than just the bottom line, sort of, if we were all going to be able to get your proprietary new reports, and if those reports are shown to hold the same validity and reliability as their data-richer forebears.

          But as you say, that is now a moot point. As things stand, with V4 you plan to give us no information and a lot less data, but you are suggesting that it’s somehow and against all logic an enhancement to your service. You don’t provide any data (or even interpolated reports…) to convince us of this improbable conclusion. You just keep repeating it. Please stop. Please honor our collective intelligence and either explain how less is more (less data, less portability), or quit trying to paint V4 that way.

  • Scott23H

    What you are asking is not practical. We can’t reserve some orders to run on a different chip in our lab.

  • Guest

    How long does a message that ‘must be approved by moderator’ take to post? In my Disquis account it states I have 1 message but fails to show a message has been processed?

  • Michael

    I am a long time customer who has tested many family members at my expense and recruited numerous new customers because I am in a related science field (genetic epidemiologist) so am excited about the product and run several genealogical related programs. In addition, I am working with 23andme to propose a new research project on genetic determinants for a particular disease that I work on using the Research Portal — so please realize that I recognize the importance of your medical research.

    However, I must say that when I look at the data regarding the new V4 chip, I am concerned with the reduced emphasis on genealogically relevant SNPs and the fact that this reduced number will make the results incompatible with Family Tree DNA and other services
    (such as Gedmatch.com). As you state bluntly earlier — your primary interest is your customers and not comparability with other company’s data, I feel I must point out how this will negatively impact you. This lost transferability would be substantial issue for individuals using your company for genealogical purposes (note that many customers don’t care to know about the medical side of their data).

    Currently, I recommend, through personal and professional circles, that testing should be done first with 23andme because I (used to) believe your service is superior for several reasons. Then, customers should transfer their data to other companies to increase odds of genealogy matches. However, if your new V4 chip data is no longer compatible nor is it as genealogically robust, I must change my recommendation and insist people test with the Family Tree DNA autosomal service. Personally, I have four kits that I purchased for family members and if they cannot be tested using the V3 chip, I would like to request a refund so that I can test elsewhere.

    I truly hope that these initial analyses of the V4 chip in regards to genealogy are not
    accurate but you have said nothing to alleviate your customer’s concerns. At a time that you are embroiled in FDA concerns isn’t the time you should be so brazen with your comments to folks concerned about loss of power in your product. As I stated, I’m a disease epidemiologist so I am very excited about the concept of 23andme, but your strength was the integration of both medical and genealogical data. One should not be undersold to advance the other. My hope is that I can move forward and work with you on my proposed
    research (presuming FDA allows additional studies to be initiated). I look forward to hearing from you regarding this concern.

  • Carl

    The additional data (not imputed) that they will be able to report is a more detailed Y Haplogroup or Mitochondrial Haplogroup (your Paternal and Maternal ancestry). They are measuring significantly more SNPs on the Y and mitochondria than they did before, so they will be able to be more precise about haplogroups.
    There may be other improvements too.

  • Scott23H

    That took me two days to moderate. Sometimes I’m faster.

    • Scott23H

      PS. I don’t approve every comment.

  • thelittlehoneybadger

    Wow … I really hope my mother’s batch gets tested with the v3 chip. As Africans, the 23andme upgrade basically does nothing for us and for now, our best bet for a detailed and more accurate admixture breakdown is GEDmatch. It would be awful if the benefit of having raw data compatible with GEDmatch was no longer a guarantee.

  • Christy Brooke Williamson

    May you pay extra for the V3 chip?

  • Scott23H

    Christy, That won’t be possible once we exhaust those chips and transition to the new chip. It would also be very difficult in the current lab process to pull out individual samples to be run on different chips.

  • Arthur

    I got my results in August and as far as I knew my results matched my expectations. However at the beginning of this month I logged into the page and noticed my results had changed drastically. I used to have only 10% of my data as Unassigned, but now I have 24.4% of my data as Unassigned. Instead of providing for a similar or more accurate DNA sequence it cut off a huge portion of my previously recorded data. Nearly a quarter of my DNA now is Unassigned. This obviously isn’t nearing accuracy but instead is truncating the previously assigned data. It would be nice if the new chip would also test at least the same amount of SNPs as before, which was the main reason why I chose 23andme in the first place.

  • me

    Where on my data do I look to see which chip was used? My sample was analyzed during the switchover – sometime during Nov/Dec so going by date is not helpful for me. The results after using 3rd party software to analyze my raw data are very different depending on which chip set I look at. Thanks!

  • Scott23H

    Amy, Your results won’t vary based on the different chips used. One way to see what chip you are on is to do a one-to-one comparison of yourself to yourself. You can do this using the gene comparison tool here: https://www.23andme.com/you/compare/

    If you are on v3, the number of SNPs will be over 900,000. If it’s v4 it will be just under 600,000.

  • Tripthi Mathew

    Has anyone tested for ancestry with V3 or V2 and with the latest version V4? Did you find any difference in the ancestry or other results? I understand the V4 chip is not enabled for testing for health reports.