Cell Counting 101 – Tip 5: How Can Variance Reveal the Precision of Your Cell Counts?

12 Comments

Tip 5 of 6

– 4 min read.

Cell Counting 101 – Tip 4 tells you all about cell viability.

What is Variance?

When performing multiple cell counts, you can calculate the variance as a measure of how much the counts deviate from the mean. Variance is used to give an indication of how similar individual cell counts are within the same sample or data point. The lower the variance, the more certain you can be that your counts reflect the actual number of cells in each sample or experimental condition.

To calculate the sample variance1, use the following steps:

  1. Calculate the mean value of the results, i.e. the sum of your cell counting results divided by the number of counts
  2. Calculate the difference from the mean by subtracting the mean from each of the cell count numbers. Calculate the square of differences and the summation of these values.
  3. To determine the variance, calculate the mean of the squared differences

When combining these three steps into one formula, you can calculate the variance by:

variance

NOTE: In the formula, x = the individual total cell count, x̅ = the mean of all the values and n = the number of values. The variance can also be denoted σ2. Here you divide by n-1 because you are calculating the sample variance, as opposed to the population variance, in which case you would divide by n.

Differences in Variance: Manual vs. Automated Cell Counting

Two factors that influence variance, are whether the cell count is done manually or using an automated cell counter, and if the concentration of cells in a sample is high or low. These differences are explained in this example 2:

cell-counting-variance-comparison-manual-vs-automated

The CV% on the y-axis represents the coefficient of variation which is equal to the square root of the variance divided by the mean. Graphs A and B show CV% as a function of the cell concentration for A) automated cell counting using the NucleoCounter® NC-202™, and B) manual cell counting using a hemocytometer. Here, the data points are scattered and only a few of them are located on the fitted curve, representing the mean. This indicates that there is a high variance due to a high deviation from the mean of individual cell counts.

Graph A presents data obtained using an automated cell counter. In this graph, the majority of the data points are located either on or close to the fitted curve (the mean), indicating that the overall variance is lower. Therefore, the mean value is close to the individual counts, revealing a lower variance.

If you compare the two graphs, it is apparent that when using an automated cell counter the variation in the measurements is lower, compared to cell counts that are performed manually. In fact, it is almost halved. Furthermore both the automated and the manual cell counting methods show an increase in CV% with lower cell concentrations. However, this increase is larger when the cells are counted manually (compare the graphs). This can be exacerbated by pipetting errors since they are more predominant at the lower and higher ends of the optimal cell concentration range with any given counting method.

Errors also occur during cell counting due to inconsistencies when determining what is recognized as a cell. If this is not very strictly defined and adhered to, variation in cell counts can occur. These factors are worth taking into consideration when aiming to obtain cell counts with a low variance.

cell counting variance comparison of manual vs automated methods

Read the next blog post that covers the importance of data validation as well as how to calculate the standard deviation. From the post, you can download a ready-to-use spreadsheet for all your cell counting calculations.

Further Reading

References

  1. SM Ross: Introduction to Probability and Statistics for Engineers and Scientists (Fifth Edition), Chapter 2, 2014. Academic Press
  2. Chemometec: Tech Note: NucleoCounter® NC-202™ Performance Data

By Christina Psaradaki, Student Assistant at ChemoMetec
Christina Psaradaki studies Human Life Science Engineering at the Technical University of Denmark. At ChemoMetec, she writes for the Cell Counting Blog.

BLOG DISCLAIMER

We welcome and encourage blog users to leave comments. ChemoMetec reserves the right to moderate all comments and delete any that could be determined as i) irrelevant; ii) containing spam; iii) containing profanity; iv) deemed offensive or; v) contain questionable content. We reserve the right to withhold any comment from being published. ChemoMetec is not responsible for the content in comments, which represent solely the views of the commenter.

12 comments

  1. Collin Dunbar |

    I’ve heard that with practice, some people are able to strongly reduce the variance of their manual counts, though I’m not sure if they all take steps to avoid biasing replicate counts based on their initial results. Are your data in Graph B from counts performed by people with high or low amounts of experience with manual cell counting?

    1. Janny Marie Peterslund | Scientific Affaris Manager, ChemoMetec |

      Dear Collin,

      In Graph B we used our automated cell count and viability instrument, NucleoCounter NC-202, which counts at a very low variance. In Graph A, we asked scientists who regularly do manual cell counting to help accumulate the data. We wanted to get a group of people who would be representative of a typical academic lab, who are often not instructed meticulously before counting, but each have their own slightly different approach to it, so we did not ask the people who do manual cell counting multiple times a day only.
      You are right that if people count manually often, they will get lower variance in their counts. But variance comes also from the pipettes you’re using (are they calibrated correctly at all times?), timing of your procedure and many other small factors, together contributing to the overall variance. This Mini Review sums up the main errors in manual cell counting that you should be aware of: https://chemometec.com/resources/mini-reviews/manual-vs-automated-cell-counting/

      Thanks for your comment – I’m adding you to the list and sending you a free magnet next week! In case you missed it, click here: https://chemometec.com/cell-counting-blog-magnet-giveaway/

      All the best,
      Janny Marie.

  2. Rod N | Quality Control |

    Given that the cell concentration can greatly influence counting results, it looks like your optimal cell concentration might be between 1-2E6 cells/ml in your comparison. Have you found there to be an optimal cell concentration overall between manual and automated cell counts?

    1. Janny Marie Peterslund | Scientific Affairs Manager, ChemoMetec |

      Dear Rod,

      Thanks for your question – a magnet is on the way next week!
      Every system has it’s own range limit and optimum. With our NC-200 shown above, for instance, we saw an optimal cell counting range of 5×10^4 to 5×10^6 cells/ml, but with the new NC-202 that has increased to 5×10^4 to 1×10^7 cells/ml. You can read this Tech Note on instrument variance to learn more: https://chemometec.com/wp-content/uploads/2021/03/Tech-note_994-2030_Cell-counting-variation.pdf

      Every hemocytometer has it’s own optimal cell counting range as well. We wrote more about that in this blog on sample dilution: https://chemometec.com/cell-counting-101-dilution-factor/

      Towards the upper and lower limits of the optimal range, cell counting results will have higher variance. And as Ray Lamb says below: Automated processes are not necessarily always accurate, but they are the same amount of inaccurate every time, and therefore are very consistent compared to manual cell counting.

      I hope this helps?
      All the best,
      Janny Marie.

  3. Ray Lamb | continuous improvement - Australian Beer Company |

    Juran places manual accuracy at 80%. Automated processes should be more consistent as the human decision making is taken out of the equation. That said, if automation gets it wrong, it always gets it wrong.
    It’s easier to tune 1000 of the same error than it is to tune 10 different ones.

    1. Janny Marie Peterslund | Scientific Affairs Manager, ChemoMetec |

      Dear Ray,

      Thanks for your comment. I agree and hope you have very consistent S. cerevisiae numbers in your brews. I’ll ship you a magnet next week – learn more, if you don’t know about our campaign already: https://chemometec.com/cell-counting-blog-magnet-giveaway/

      All the best,
      Janny Marie.

  4. David Laursen | Aarhus University |

    Cool magnets you’ve made 🤓

    1. Janny Marie Peterslund | Scientific Affairs Manager, ChemoMetec |

      Hi David,
      Thanks a lot – we really like them, too! I’m Contacting you shortly to find a way to ship them to you.
      Many greetings,
      Janny Marie.

  5. Celine | Project manager |

    I love lab magnets ! I have a whole collection on my fridge. Happy to see yours joining us soon.

    1. Christina Psaradaki | Student Assistant, ChemoMetec |

      Hi again Celine,
      You’ve secured yourself half of the magnet series (2/4)! I’m sure they’ll be a great addition to your collection.
      Kindly,
      Christina

  6. Guillaume Grosjean | Research Technician - Nancy (France) |

    Interesting this article, and thanks for magnets.

    1. Janny Marie Peterslund | Scientific Affairs Manage, ChemoMetec |

      Hi Guillaume,
      Bonne Année and I apologize for the late reply! I’ve sent you an email so I can ship you some magnets. It’s a late Xmas present – yay!
      Best wishes for 2022,
      JM.

Leave a comment

Your email address will not be published. Required fields are marked *