## Upcoming Events

## Applied & Computational Mathematics seminar: Recent Advances in PINNs and Deep Neural Operators.

Sep 20, 2024, 10:00 - 11:30 AM

## CAGS: Minkowski rings of polytopes and power closed ideals

Sep 20, 2024, 12:30 - 1:30 PM

# Applied & Computational Mathematics seminar: Separating Flow Cytometry Populations: A Probabilistic Approach

Nov 11, 2022, 1:30 - 2:30 PM

**Speaker:** Danielle Middlebrooks,NIST

**Title: **Separating Flow Cytometry Populations: A Probabilistic Approach

**Abstract:** Flow cytometry is a widely used technique for single-cell and particle analysis with applications in immunology, cancer biology and infectious disease monitoring. As the complexity of FC experiments increases (with number of cells/particles and measured characteristics), traditional flow cytometry methods for distinguishing populations (e.g., gating) has become impractical due being both subjective and time-consuming to the user. We developed an objective method to distinguish populations described by corresponding probability density functions (PDFs) while avoiding gating. We begin with two samples, one having a single cell type and the other contaminated with an unknown amount of a second type. We represent their probability densities Q(x) and P(x) using a set of orthogonal basis functions; this allows us quantify uncertainties associated with having finite data. Once the PDFs are constructed, we answer the following question: How much of Q(x) can we subtract from P(x) to find the fraction of the remaining unknown second cell type? To answer this, we formulate a constrained optimization problem that utilizes non-negativity of PDFs and the assumption that our constructed PDFs will have partially disjoint supports. We demonstrate how our method can efficiently distinguish populations and remove initial subjectivity in the data analysis while quantifying the uncertainty.

**Time:** Friday, November 11, 2022, 1:30pm-2:30pm

**Place:** Exploratory Hall, Room 4106 or **Zoom**