At Ken Analytics, we’re committed to applying advanced statistical techniques to solve real-world problems. We take pride in the diverse range of projects we’ve tackled for clients and collaborators across various disciplines. Below, we highlight a few of our past projects, offering a glimpse into how our work has helped clients make informed decisions, optimize their processes, and navigate complex problems with confidence.
Improving RNA Seq Analysis: Reducing False Positives when looking for Differentially Expressed Genes
When studying genetic activity, researchers are often focused on identifying which genes change their RNA production in response to different conditions. For example, they may want to understand which genes control a tomato’s reaction to drought or which genes are more active in cancer patients compared to healthy individuals. The ability to identify these “differentially expressed” genes is crucial, but traditional methods have overlooked a key source of uncertainty—model uncertainty. As a result, these methods often underestimate the risk of false positives, where genes are incorrectly identified as significant when they are not.
With the advent of next-generation sequencing technologies, which can measure the expression of tens of thousands of genes at once, the potential for false discoveries has increased. If researchers are overly confident in their results, they risk wasting valuable time and resources investigating genes that don’t truly have an effect.
During his time at Iowa State University, Steve developed a quasi-likelihood method that addresses previously overlooked uncertainty in the mean-variance relationship of count distributions. This method allows researchers to more effectively control the False Positive Rate (FPR) when selecting genes for further investigation, without losing the power to detect truly important genes.
To make this approach more accessible, Steve developed an R package that implements this methodology, which has since been widely adopted in the field. The publication detailing this work has been cited hundreds of times, reflecting its significant impact on improving the reliability of gene expression studies.
At Ken Analytics, we’re committed to applying advanced statistical techniques to solve real-world problems. If you’re facing similar challenges in controlling false discovery rates or appropriately accounting for modeling uncertainty, our team is ready to help with innovative, statistically sound solutions.
Improving Flow Cytometry: A New Method for Evaluating Measurement Error
Flow cytometers are powerful tools that can measure the expression of tens of thousands of cells per second, providing valuable insights in fields like immunology, oncology, and cell biology. However, once a cell is measured, it becomes mixed with all the other cells, making it impossible to directly assess how repeatable or reliable the measurement is for an individual cell. This uncertainty presents a challenge for researchers who need to ensure the accuracy of their data.
To address this issue, we developed an innovative approach using flow sorters. Flow sorters allow users to set a threshold, separating cells based on their measurement. Cells that fall below the threshold are directed to one container, while those above the threshold are sent to another. By remeasuring the cells in each of these containers after sorting, we can assess the measurement variability around the threshold.
For example, if there were no measurement error, we’d expect to see values from the “below-threshold” group range up to the threshold but not exceed it. The sharpness of this cutoff—how distinct the separation is in the second measurement run—reveals the level of measurement variability. This method provides a way to directly assess measurement error and improve the reliability of flow cytometry data.
To make this approach more accessible, we’ve provided code for clients to implement this analysis in their own work. By applying this method, researchers can gain more confidence in their flow cytometry results, ultimately leading to better-informed conclusions.
If you need assistance improving measurement accuracy, our team is here to help with tailored statistical solutions.
Improving Cell Counting Accuracy: A New Standard for Evaluating Methods
For over a century, manual cell counting under a microscope has been the gold standard for determining cell concentration in a solution. Researchers would count cells in a small sample and then extrapolate the result to the entire volume. While this method has been effective, it is labor-intensive and highly prone to measurement errors due to the small sample size.
As automated cell counting instruments entered the market, researchers faced the challenge of determining whether these new methods could reliably replace manual counting. Traditionally, many researchers treated manual counts as the “gold standard” and evaluated automated methods based on how closely they matched the manual results. The problem with this approach is that no automated method could ever surpass the perceived accuracy of manual counting, even if the automated methods were, in fact, more accurate.
We developed an innovative solution. By performing serial dilutions and comparing the measured cell concentrations to the expected values based on known dilution factors, we were able to assess the quality of both manual and automated methods. This allowed us to evaluate manual counting with the same rigorous criteria applied to automated methods, leading to fairer comparisons and more informed decisions about which method to use.
The analytical approaches we developed were so impactful that they were eventually adopted as an official ISO standard. This work not only improves the accuracy of cell counting but also helps researchers make better decisions when selecting counting methods, ultimately saving time, resources, and improving results.
If you’re facing similar challenges in your research or need help evaluating methods, our team is ready to assist with cutting-edge, reliable solutions.
Ken Analytics 