Resurrecting a Seahorse XF96

Those familiar with Seahorse XF instruments are aware that since the company was bought out by Agilent, there has been a tendency toward “sunsetting” older instruments and encouraging users to upgrade to newer models. Most recently, this came in the form of an email claiming that, not only would the older XF96 instrument be no longer supported, but also the availability of consumables would cease…

Naturally, Agilent’s preferred solution to this problem is for the end-user to drop $100k+ on a newer XFe96 or XFpro machine. For those like myself without deep pockets, this is not an option. We’ve used Seahorse products for a long time (in-fact we had one of the very first XF24 devices sold), but we’re not in a position to drop that kind of money every 5-6 years just to keep doing our experiments.


Crappy PCs

Adding to the problem of the instrument itself becoming obsolete, the XF96 shipped with a “POS” terminal – a piece of shit touchscreen 32-bit PC running Windows 7 with a measly 2GB of RAM. The touchscreen interface never lined up properly, and simply added to the overall slowness of the machine (here’s the underside of the offending PC – yes those are RS232 serial ports).

In addition to frequent crashes (sometimes during the middle of experiments!) many universities including my own will not allow older PCs such as this to connect to their network, so this effectively renders the PC incapable of receiving security upgrades.The use of USB thumb-drives to transfer data on/off the PC is therefore risky.


Simple – Just Swap in a New PC!

You’d be mistaken for thinking the solution here is to simply swap in a brand-new PC and transfer the software. That’s because there are a number of software and hardware dependencies which anchor the XF to the VERY specific PC it ships with…

  • The XF communicates with the PC via two separate cables: an old-school RS232 serial port for the instrument itself, and a separate USB/serial port adapter for the barcode reader.
  • Multiple subroutines within the XF software need old-school plug-ins and dependencies. This includes Adobe Flashplayer, MS Edge plugins, Visual C++, and running most things in compatibility mode for Win7.
  • The system needs a VERY specific version of Microsoft Excel 2010.

In-fact, in addition to some minor tweaks regarding temperature control inside the instrument, one of the key advantages for the XFe96 versus the XF96 as marketed by the company, was the ability to run on a modern 64-bit Win10 PC. In effect, they crippled the XF96 so badly you couldn’t even upgrade the PC connected to it!


Sounds Like a Challenge

Naturally, not wanting to continue running our XF96 on a defunct old PC, I undertook the challenge to get a modern Win11 PC speaking to it. The PC I chose was this cheap fanless mini-PC that sticks to the back of a monitor. It comes with Win11 preinstalled and 6 USB ports, so plenty of room for expansion. Since it doesn’t have any serial ports, I also picked up another USB/serial adapter cable.

Below are the steps to install the relevant software. Note that many of the files are no longer available through official channels such as Microsoft, so you need to find a link (e.g., from an old discussion forum) and then copy the URL into the Wayback Machine, then visit a cached version of the old link. It goes without saying this is all at your own risk, with no implication here that any of the linked resources are reliable, virus-free, or whatever. There’s also some light registry editing involved, which comes with a FAFO disclaimer!

  • First install Microsoft Visual C++ redistributable 2010. Even though most new PCs are 64-bit, you will need both the 64-bit and 32-bit (x86) versions of this. I obtained these files here, but you can also find them using the cached link method as described above.
  • Download and install Java.
  • Download and install Microsoft Edge WebView Runtime X64. This should come via Windows Update on a modern PC, and it replaces the Active X module in Internet Explorer that was required on the original Seahorse PC.
  • Install the XF96 installer package. The file is no longer available from the Agilent website, but the one you’re looking for is “XF96SetUp.exe” and the Software version is from 2018. DO NOT install the “Wave” software, as this will not play nice with the XF96. Hopefully any lab with an existing XF96 machine should have a copy of the software installer sitting around.
  • Install Microsoft Excel 2010. If missing the original DVD, you will need to find an ISO or EXE file to download. You will also need a Key Extractor to recover the software key for Excel from the old PC (see below). This site lists all the links for Office downloads, and the file you’re looking for X16-30329.exe from this link (need to use Wayback to find a cached version). **You can’t use any Office 2010 ISO file. It has to be from the retail version of Office Pro, NOT the pro-plus version which was for enterprise distributions. If you install the wrong version, the recovered software key will not work.
  • Download and install the driver for the original USB/serial cable that came with the XF96. This should be a Gigaware device, and will install the file “Svk2pl64.sys” in the folder C:\Windows\System32\Drivers\ You may also need another driver for the additional USB/serial adaptor purchased to go with the new PC.
  • Extracting the software key from the old copy of Excel is not something Microsoft wants, so key-extractor software tends to reside in dark corners of the internet.  Tread carefully – such software WILL trigger a virus warning and quarantine on a modern PC, so download and transfer it to the old PC as a .zip file and then only unpack it on the old PC. Remarkably, in my case doing this did not trigger a virus warning on the old PC – a good sign that it’s not suitably protected!



You’d think the simple approach would be to use the old black Gigaware cable for the barcode reader on the old PC, to connect the barcode reader on the new PC, then use the fancy new USB cable to replace the serial/serial connection. NOPE – for some bizarre reason this doesn’t work (the barcode reader will connect, but won’t capture anything). The new cable must be used for the barcode reader.

  • Start with all cables disconnected. Plug the Gigaware cable into a USB port. Type “DEV” in the Start Menu and open “Device Manager”. Under “Ports”, double click on “Gigaware USB to Serial Cable”. Under the “Port Settings” tab, click “Advanced, and set the port to COM1.
  • Connect the other (serial) end of the cable to the XF port labeled COM1 (looking from the back of the machine this will be on the right).
  • Now turn on the XF96. Launch the XF Utility App (located in C:\Program Files (x86)\Seahorse Bioscience\XFReader96\Utility\). Under the “Connections” tab, select COM1, and click connect. Successful connection yields a green light. If prompted for a passcode, use “OSTER”.
  • With the utility app still running in the background, now connect the second (new) USB/serial cable to a different USB port. Open Device Manager again, and use the same procedure as above to assign this cable as COM2.
  • Then connect the other end of this cable to the port labeled COM2 on the XF (looking from the back of the machine this will be on the left).
  • Finally, in the Utility App, in the “Barcode” tab, select COM2 and connect to the barcode reader.



Every Seahorse XF machine is slightly different, so a bunch of instrument-specific settings are stored in the registry for call-up by the software. This includes the offsets for temperature regulation, and the precise positions of the motors for different functions. All these settings have to be transcribed over to the new PC. There are 3 different ways of doing this…

(1) Before closing out the original PC, export a copy of the registry settings containing all the parameters from the old machine. These are stored in HKLM\SOFTWARE\WOW6432Node\Seahorse Bioscience\XF96_CTRL

(2) Alternatively, before disconnecting the old machine, launch the Utility App and take screenshots of all the tabs (especially Motors). Or just write down the motor positions and other important numbers from the Utility.

(3) A third approach is in the main Seahorse software itself – navigate to the “Instrument Setup” screen and click the “Administrator” button (right side). On the left, click “Hardware Settings” – a password prompt will appear (PW: GENIUS). Then click the various tabs to note down important parameters such as probe positions.

Once you have all the settings noted, keep them in a safe place. Then open the Utility App or the main Seahorse software on the new PC and enter all the values. Or just edit the registry. It is essential at this point to restart both the XF machine and the PC, and make sure all the settings have “stuck” in the registry.

*Oddly, when set-up was all finished, the Seahorse software would only recognise that there were 2 injection ports, not the usual 4.  There’s a registry setting “NumPorts” that was set to 2 – set it to 4 (decimal) and restart, then all the ports will be available.


IT’S ALIVE! (but what about consumables?)

If you’ve done everything correctly up to this point, you should now be able to turn on both the PC and the XF, and have them play nice with each other. At that point, it’s time for an experiment…

BUT… recall the original email from Agilent said that XF96 comsumables would no longer be available? This is not an immediate concern, because XF consumables never really go off (we’ve used cartridges 5 years past their nominal expiration dates with no problems). But, for those in need of new plates and cartridges, it appears the claim they are not available is simply untrue at this time (October 2023)…

There are two types of consumables on the Agilent website. This type (catalog #103798-100) is compatible with both XFe96 and XFPro machines, and we have found these cartridges and plates work perfectly well inside an old XF96 machine! There is also a newer line of plates (103777-100) that is ONLY compatible with the newer XFPro machine, and these won’t even work with the XFe96. Don’t buy those ones.

It’s notable that very recently the XFe has also disappeared from the Agilent website, so one imagines they are planning to phase that out soon as well. But, in the mean-time, it appears that plates/cartridges marked as XFe96 compatible will work on the XF96.

It should also be noted there are additional new plate types (e.g., spheroid/organoid plates), wherein the plate itself is a completely different shape, and these physically will not fit inside the old machine. Apparently you can CNC-machine the metal block plate holder inside the machine so these new plates will fit, but I wouldn’t know anything about such shenanigans!


It’s very possible to keep a ~10 year old Seahorse XF96 functioning, and have it talk to a modern Win11 64-bit PC, with all the advantages that conveys (virus protection, software and OS updates, networking, watching YouTube while your experiment runs).

The relentless march of planned obsolescence in scientific apparatus shows no signs of slowing, so I have no idea how long this particular set of hacks will buy us. Maybe a year, maybe more? Either way, for those without $100k+ to spare this is really the only option.

P.S. Used XF96’s can be had on eBay for $2-3k, so in theory with a cheap mini-PC you could build a working Seahorse system for well under 5 grand!

    Spring ’23 Update

    Things have been trucking along in the lab’ since last fall…

    • Anya Wang (Masters Student) graduated in May, and will defend her masters’ thesis in June.
    • Caio Tabata Fukushima (Undergrad Assistant) also graduated in May, and will return to URMC in the fall as a PhD student in the Pharm/Phys graduate program.
    • Several papers published nominally during 2022 finally made it to journal pages with a 2023 publication dates (see here).
    • Paul will be presenting Caio’s work on reverse electron transfer in mitochondrial complex I, at the American Heart Association Basic Cardiovascular Science meeting (AHA BCVS) at the end of July in Boston.
    • We still have an open post-doc’ position (link) and have interviewed several candidates but unfortunately visa and other issues prevented us making an hire.
    • Paul’s work in the area of scientific integrity continues apace, with over 50 papers reported on last year, plus numerous others found to contain problems during peer review. I’ve been using ImageTwin a lot for pre-screening, and frankly it’s now annoying to be sent a paper to review that hasn’t gone through some kind of similar tool. For those interested, I now have an annotated database with almost 1000 verified examples of image manipulation across more than 500 papers, for use as a training set for AI models or simply as a teaching tool for research integrity programs.
    • Paul participated in Skype-A-Scientist for the 4th year in a row, this time talking about mitochondria with 7th graders in Kentucky

    Cashing in on Oxaloacetate

    I’ve written previously about the pseudo-science underlying the promotion of oxaloacetate (OAA, yes the Krebs’ cycle intermediate) as a dietary supplement. The TL/DR, is that the company involved (Terra Biological) touted OAA as a treatment for various diseases, based on model organism data (worms and mice), which earned them a warning letter from the FDA. The “evidence” was published in predatory journals (many of which don’t exist any more) with minimal conflict-of-interest disclosures. Couple with the ridiculous doses you’d need in humans to achieve the same level as used in animals, and a number of fundamental misunderstandings about the basic biochemistry of OAA, and you can see why I rightfully called it “shenanigans”.

    Shortly after that post was published, the company’s CEO Alan Cash called me on the ‘phone, threatening to sue if I didn’t take it down. I declined and asked him to point out exactly what in the post was untrue or would form the basis for a lawsuit. I never heard back from his lawyers.

    More OAA Garbage – now with added “Clinical Trials”

    Recently, I was alerted to a new paper reporting on a clinical trial (I use the term loosely) for the use of OAA as a therapy for PMS. The senior author is Alan Cash, and let’s just say there are a number of issues with this paper, which should give pause to anyone considering wasting their money on Jubilance for PMS…

    (1) The study allegedly took place at the “Energy Medicine Institute” in Boulder CO. A quick search shows that the website has lapsed and been taken over for advertising. The business address (27 Arrowleaf Ct., Boulder, CO 80304) shows up on Google maps as a house in a residential neighborhood. The business is registered under an SIC code (Standard Industrial Classification system) of 7311, which is advertising agency. There are several other organizations with similar names (e.g., Energy Medicine Research Institute and Energy Medicine Institute), but the lead author on the paper is Lisa Tully PhD, who appears to have died recently in Boulder CO, so I guess that tells us which one we’re dealing with.

    (2) The only MD on the paper is John Humiston MD from the “Center for Health and Wellbeing” (no address or location given). There’s a John Humiston who, according to QuackWatch is known for injecting urine into patients in Mexico. Quackwatch says he’s on the staff of the San Diego Center for Health and Wellbeing, which is located near Terra Biological HQ. Notably Humiston is not listed on their site, but this does appear to be a medical facility so it’s possibly where the study actually took place.

    (3) The methods section states “Institutional Review Board approval was obtained”. Academic medical institutions have their own IRBs, but others have to go to a commercial IRB. In this case, it appears they used ICRM, which seems to be a preferred venue for quack stem-cell therapies. It’s not very professionally run (e.g., the company blog includes a story about the Mayo Clinic… in Rochester NY!) Their client portal and online submission system is “coming soon”. Notably, nowhere in the paper is it stated how the patients were recruited.

    (4) Looking at the part A vs. part B of the study (Table 2) they say the reason they did B is because there was a “carry over” effect in part A, where the patients who got the drug first ended up showing significant improvement even if they were later on the placebo. What’s interesting is the 4 scores in the placebo group were 25.9, 21.4, 24.0, and 6.5. The funny thing is, when they did the study the 2nd time, the numbers were 28.1, 15.0, 15.1, and 6.2 in the placebo group. The only thing that really changed was the p values came down. So, the “significant effect” in the placebo group actually became even more significant when they received the placebo first and then the drug.  This indicates there was a massive placebo effect. The second study (part B) was designed to “overcome the carry over effect”, but all it did was prove that the reason they saw an effect in the placebo group was NOT due to carry over!

    (5) There are several typos indicating these folks have no clue what they’re talking about when it comes to basic research (and the fact these were not picked up during peer review tells us about the quality of editing by the journal). For example in the discussion they refer to “C75B1/6” mice. WTF?

    (6) The statistical methods are weird. Looking at one of the scores (Becks Depression) in Table 1, they state the baseline errors as standard deviation (SD, 27.8 +/- 11.3). But when it comes to the results (Table 2), they show the numbers in percentages and they give standard errors (SE). The SE is the standard deviation divided by the square root of the number of samples. In part A they had 26 patients, so we you can back calculate from their SE value to get a SD. Doing that, the percent change is not 52.2 +/- 8.8% as they have it listed in Table 2, it’s actually 52.2 +/- 41.3% (using SD instead of SE). This indicates the effects were highly variable. Using these numbers to convert the percent changes back into real units, we can calculate the actual Becks Depression score for both groups. At baseline it was 27.8+/-11.3 as they state (mean +/- SD). For the OAA group it was 42.3+/-33.5. Further calculating the 95% confidence intervals, the baseline group would range between 23.5 and 32.1, while the OAA group would be 28.3 to 56.3. As such, there’s no way these are significant by ANOVA. That probably explains why they chose not to use ANOVA to analyze the data.

    (7) Lastly, why is this buried in an obscure Korean journal without an official impact factor? The journal recommends but does not require a data sharing statement and policy, but there is none for this paper.

    But there’s more… OAA and COVID!

    Basking in the glory of a clinical trial, it was only a matter of time before the COVID bandwagon would be jumped on. Thus, we find ourselves in the unfortunate situation of viewing this paper, reporting on another clinical trial for the use of OAA to treat chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME) and long-COVID fatigue.

    In general, CFS/ME is a very poorly defined and poorly understood set of disease symptoms, for which there is currently no cure. As such, people suffering from this syndrome are often willing to try anything, including $500 a bottle supplements (FYI, PubChem lists 62 suppliers of OAA, for those interested).

    In terms of what’s wrong with this paper, an anonymous commenter on PubPeer has already raised numerous issues, including undisclosed conflicts of interest, deviation from pre-registered trail endpoints, missing data, and several other problems. The author of the PubPeer critique cites my previous post, agreeing that many of the claims made regarding OAA as a therapeutic are at odds with basic biochemistry principles.

    There are additional problems to flag with this study:

    (1) The paper lists the affiliation of its senior author David Lyons Kaufman, MD, as the “Center for Complex Diseases, Seattle WA, USA.”  Such a business with a person of the same name does really exist, but it’s based in Mountain View CA, not Seattle.

    There’s a David L. Kaufman in Seattle, who appears to be affiliated with the “International Peptide Society.” This organization appears to promote the use of “peptide therapeutics” without ever stating what the peptides are. They offer all kinds of educational events and certification in the use of “peptides” in regenerative therapy, but their website is devoid of ANY peer-reviewed scientific literature. Their “Find a Practitioner” page is a laundry list of new age quackery – seriously, do not jump down that rabbit hole!

    (2) The National Clinical Trials Database page for the CFS/ME trial lists our old friend Lisa Tully as the Principal Investigator, and the Energy Medicine Institute only location for the trial. David Kaufman is not listed anywhere on the NCT page.  Surely a big change such as the lead investigator in Colorado dying, and then moving the trial site to Seattle (or is it Mountain View) would warrant an update to the NCT database?  As such, it’s not exactly clear where this trial took place.

    (3) More company PR about the use of OAA for CFS/ME claims that the reason Kaufman got interested in OAA is he saw it was on a list of metabolites that declined in CFS/ME patients, in a 2016 metabolomics study. The problem is, a detailed look at the data reveals that the decline in OAA was not statistically significant after correcting for false discovery rate (Q value vs. P value).  The actual data are on the left (red is controls, blue/green is CFS patients), showing essentially no effect.  OAA didn’t even make it onto the bottom of the chart for “relative importance” of metabolite changes (Figure 3 of the paper). Another metabolomics study of CFS (published in a better journal) did not find anything relating to OAA. As such, there is no credible evidence that OAA levels are lower in CFS/ME patients, so the entire premise for the study falls flat.

    (4) In that same PR piece, it’s notable that Kaufman is talking about results from ongoing trials in CFS/ME patients in October of 2021. The published study claims  recruitment began in February 2021, and the two sets of patient characteristics have some striking similarities:

    The PR post states 52 patients across 2 different doses (500mg or 1000mg twice daily), average age 49, 77% women, and an average 25% drop in fatigue score at 6 weeks.   The published study reports 76 patients, addition of a 3rd dose group (1000mg 3x daily), average age 47, 74% women, and a reduction in fatigue of 22.5% to 27.9%.

    It is therefore reasonable to assume the patients being discussed in the October 2021 PR piece have significant overlap with those in the final published paper in June 2022. The only addition for the paper is the long COVID group, which makes the final sentence of the PR piece particularly interesting…

    “OAA studies are being pursued in Alzheimer’s, ALS, myasthenia gravis, cancer, ME/CFS and long COVID. The long-COVID double-armed, randomized, double-blinded placebo-controlled trial is being funded by the same company (Terra Biological) that provided the supplement to Kaufman.”

    Clearly the long COVID study that ended up being published here was not double-blinded or placebo-controlled. Clearly the statement on the NCT database page that the study would be double-blinded and placebo-controlled, was not followed through on. The fact that the entire NCT page for this published study is about COVID patients, and doesn’t actually mention non-COVID related CFS/ME at all, is also a problem. There does appear to be another clinical trial registered for OAA in CFS/ME, but it only just started recruiting patients. The CFS/ME patients who took part in the trial reported here, were outside the boundaries of the NCT registered trial for long COVID patients. Treating patients first and then registering the trial later, is definitely not kosher!

    Despite several red flags regarding these trials, news outlets continue to regurgitate the company’s PR claims. Terra Biological continues to dance at the boundaries of actual rigorous clinical trials and medicine, by doing things that look sort-of like trials to the genera public, while collaborating with weird quackery institutes, dosing patients outside the boundaries of what they register with the NCT database, and then publishing the data in journals that don’t seem to give a crap about rigor.

    As I have stated before and will do so again – run, don’t walk, away!


    Talking with Brandon Berry over the weekend, the subject of aging “Hallmarks” came up. I’m not a fan of this terminology, whether it’s applied to aging, cancer or any other biological condition. Surely there’s got to be something more appropriate?

    What does “hallmark” mean?
    The etymology of “Hallmark” imbues a very specific meaning – uniqueness.  In Ye Olden Days, crafts-persons would set up a Guild, a sort-of business association to protect their products and services. The Guild would sometimes be housed in a Guild Hall (an important building in the town, often sharing premises with the town hall or council offices). The Guild Hall would allow members of the Guild to use the Hall Mark – a unique stamp – to mark their wares, so that buyers could be assured of authenticity.
    (img from

    In theory, using a hallmark ALONE made it possible to tell if what you’re looking at is the real deal. Is this a real 24-karat gold ring made by a master jeweler, or a cheap knock-off?  The sole purpose of the hallmark is authentication.

    Why “hallmark” should not apply to biology
    In biology, long ago someone decided that “hallmark” means “common characteristic”, which is a completely different meaning. Now, instead of a hallmark being a unique identifier (like a fingerprint), it’s just a characteristic found a lot of the time (like blond hair). The problem is NONE of the so-called “hallmarks” of aging or cancer are unique to those conditions. They can all be found in other conditions that are not aging or cancer.

    Look at aging…

    (from 10.1016/j.cell.2013.05.039)

    Mitochondrial dysfunction? Everywhere!  Epigenetic Alterations? All over the place. Altered cell communication? Heard them on Spotify last week. None of these things are unique to aging, and frequently lots of them are found together in situations that are not aging.

    What about cancer…

    (from 10.1016/j.cell.2011.02.013)

    Resisting cell death?  Hello drugs.  Inducing angiogenesis? Meet hypoxia. Evading growth suppression… that’ll be development. Even taking hallmarks into a more specific field such as metabolism doesn’t solve the problem, as many of the phenomena (e.g., the Warburg effect) are found across various other biological states.  The problem is with the term hallmarks itself.

    So what’s a better term?
    I’ll concede that the more hallmarks one finds in a given condition, the more likely it is you can label the condition as bona fide aging or cancer. In-fact, one could even say it is ONLY possible to label something as aging or cancer if ALL of the hallmarks are present (and this conveniently ignores that we may not have identified all the hallmarks yet).

    Hmm… everything has to be there for it to be complete. You know what that sounds like?   HORCRUXES!

    For the uninitiated, in the Harry Potter novels, horcruxes are physical objects used to store a person’s soul. The dark wizard Voldemort disperses his soul across 6 horcruxes plus his body, so the only way to kill him is to destroy all 7.

    If the definition of aging or cancer ABSOLUTELY requires a complete complement of hallmarks, they’re not hallmarks. They’re horcruxes. To destroy the thing (aging, cancer etc), you have to fix/destroy all the horcruxes.

    I would argue that horcrux is a more appropriate term than hallmark when applied to definitions of biological states. It acknowledges that multiple such entities are required to meet the definition of the disease. It allows for a condition in which horcruxes can be present in other settings that are unrelated (e.g. Harry Potter didn’t realize until near the end that he himself was being used as a horcrux by Voldemort). It also acknowledges (as in the case of HP) that part of the problem is just figuring out what the horcruxes are, and we’re a long way off from being able to say we’ve found all of them. My personal opinion is that it sounds better… “stabbing the 7 horcruxes of aging with a basilisk fang!”  And finally, it acknowledges a certain degree of magical thinking has to occur, to believe any complex biological process can be distilled into a half dozen simple things.

    I don’t know quite how the decision was made to wholesale change the definition of hallmark from “unique identifier” to “commonly found characteristic”, but my favorite conspiracy theory is the drafters of the original papers were big fans of the Hallmark TV channel  😉

    Spring updates (and we’re hiring)

    Since last fall, some big changes around here…

    • Chaitanya Kulkarni, former post-doc’ in the lab, has moved to industry and is now with Rheos Pharmaceuticals in Boston.
    • Our former lab tech’ Alyssa Tavino also left in December to go to back to school to be a Physician’s Assistant.
    • Our long-standing NIH grant (R01-HL071158) received a 1st percentile score at November study section, and we are now in possession of a Notice-of-Award, meaning the lab is funded for the next 4 years. Yay!

    So, at the moment the lab is somewhat running on a skeleton staff.  We are in need of new post-doctoral fellow (job ad is here) and a new lab’ technician (position is listed here, please search for ref #233866, as  the system won’t let me provide a direct link).  Both positions are funded by the NIH grant linked above, for those wanting a better idea of the scope of work that the positions will entail.