Cell count accuracy

[Gahr]

I recently started doing cell counts to try and check the viability of my yeast and to see if I am actually pitching the amount of yeast I am aiming for. I use Jamil's calculator to determine the size of my starters, and so far I it looks like I get what I'm aiming for (Let it be known: I have never doubted the pope!). However, I have noticed that I probably need to do multiple counts on each starter, since my results can vary with as much as 25 % from one count to another. My question then is, am I doing something wrong since I get this variation, or should I just do a certain number of counts and use an average to calculate the total number of cells? I would really like to start keeping my own yeast cultures on slants, but I figure I need to learn hoow to do proper cell counts before I even think of building my own yeast bank...

[ChrisKennedy]

There is a huge amount of error introduced by the large amount of dilution you have to do in order to count cells easily on a hemocytometer. You may be diluting slurry 100x to get a readable slide. This leads to a lot of variation if you aren't extremely careful with your measurements.

But if you are counting the starter and not a slurry, there should be less variation simply because you have to dilute it much less. The only thing I can say is be more careful with your measurements or buy better tools for measuring out slurry/diluting the slurry.

[Gahr]

I have been diluting 100x when counting cells in my starters. Should I dilute less? What would you suggest? 1:10? 1:20?

[ajdelange]

The variation in what you count, as a percentage of what you count, is inversely proportional to the square root of what you count. If you count only ten cells you can expect a variation of 100*SQRT(1/10) = 31.6%. If you count 100 cells its 100*sqrt(1/100) = 10% and so on. Now you can count them 10 at a time (and average) with a high dilution or 100 at a time with a lesser dilution. What's important is how many you count. For 1% you would need to count a total of 10,000.

This is based on the assumption that dilutions are made from a well mixed starter or wort and that samples are drawn from a well mixed dilution. The degree of dilution in not a factor. If you dilute more you will have to take more readings and if you don't dilute enough there will be so many cells that you can't keep track of them all during counting but other than that one dilution rate is no more error prone than another. I'd say dilute until you have around 250 cells in the ruled area (10 per large square). That should be manageable to count and would give you an expected variation of 6%. If you want 3% count that dilution 4 times and average.

[Gahr]

Thanks a lot!

[ChrisKennedy]

The variation in what you count, as a percentage of what you count, is inversely proportional to what you count. If you count only ten cells you can expect a variation of 100*SQRT(1/10) = 31.6%. If you count 100 cells its 100*sqrt(1/100) = 10% and so on. Now you can count them 10 at a time (and average) with a high dilution or 100 at a time with a lesser dilution. What's important is how many you count. For 1% you would need to count a total of 10,000.

This is based on the assumption that dilutions are made from a well mixed starter or wort and that samples are drawn from a well mixed dilution. The degree of dilution in not a factor. If you dilute more you will have to take more readings and if you don't dilute enough there will be so many cells that you can't keep track of them all during counting but other than that one dilution rate is no more error prone than another. I'd say dilute until you have around 250 cells in the ruled area (10 per large square). That should be manageable to count and would give you an expected variation of 6%. If you want 3% count that dilution 4 times and average.

Are you assuming perfect measurements or is that factored into the variation you are speaking of?

[ajdelange]

In using a hemacytometer a small volume, Vs, (it looks to be about 3mm x 3mm x 0.1 mm - this last number is labeled on my unit) is drawn from the culture and the number of cells in that small volume counted. The numbers I threw out are based on the assumption that any cell is equally likely to wind up in any volume Vs regardless of where that volume is drawn from the sample i.e. that the sample is uniformly mixed. If the sample is diluted and there is an error in the measurements associated with the dilution then that error will appear in the final answer. e.g. if you count 100 cells thinking the dilution factor is 10 when it is actually 10.1 that will introduce a 1% error. But the unceratainty from a count of 100 cells, if we can assume perfectly uniform mixing, is sqrt(1/100) = 10% i.e. larger than a 1% dilution error.

But to try to answer your question diretly, yes, the numbers I gave speak only to the normal statistical variation associated with sampling a population. Uncertainty in dilution adds to the uncertainty though it appears that the statistical variance would be larger in many cases.

[drummstikk]

my results can vary with as much as 25 % from one count to another

The sample variance adjelange described should tell you the statistical limits of your accuracy. You may already be on top of this, but there are a few things you can do to improve the biological accuracy.

Like ChrisKennedy pointed out, it's really important to get the dilution right with a reliable pipette. But if you have a crappy one, you can reduce your error by making several small dilutions. So, instead of diluting 10 uL in 1 mL to get a 100x dilution, dilute 1 mL into 2 mL seven times for a serial dilution of 128x. You're just repeating a small dilution over and over until you get the large dilution you're looking for. You'll have to use more of your slurry (1 mL vs. 10 uL), but it will be worth it for the accuracy.

The other thing is to use one of those push-button counters -- they look sort of like what an umpire uses to keep track of balls and strikes.



If you don't have one of these, I can tell you, they make such a huge difference.

And, you might try following a pattern of counting. I have never really done this, but then again, I haven't been as careful about my counts as you are.

I figure I need to learn hoow to do proper cell counts before I even think of building my own yeast bank...

You might surprised how well things will work even if you don't do counts. If you step up the yeast the same way every time -- the same ratio of volumes (plate to 10 mL to 100 mL to 1 L is what I do), the same temperature and agitation, and the same time at each stage, then I bet you won't be able to taste whatever variation there is in your pitching rate.

Now, I have never measured as accurately as you are, and I will be very eager to see what you find!