# Robust and scalable Bayesian analysis tools for single cell CRISPR screens with sequencing- and imaging-based readouts

> **NIH NIH F31** · HARVARD MEDICAL SCHOOL · 2024 · $43,203

## Abstract

Project Summary
Single-cell CRISPR screens enable genetic perturbation studies of the functional genome at unprecedented
scale and resolution. There are two major types of single cell CRISPR screens: sequencing-based screens and
imaging-based screens. Sequencing-based screens measure the impact of CRISPR-induced genetic
perturbations on phenotypes which can be read out via single cell sequencing, such as transcriptome-wide gene
expression or chromatin accessibility. Imaging-based screens profile a range of image-based cellular phenotypes
using fluorescence microscopy, including cell morphology and the localization of fluorescently-tagged proteins.
Together, these two types of single-cell CRISPR screens can study diverse cellular behaviors which were
previously inaccessible to pooled genetic screens. However, the datasets generated by these screens are large
and complex, requiring fast and accurate computational analysis tools to interpret the phenotypic effects of each
genetic perturbation. Yet existing methods are often not statistically robust or are otherwise prohibitively slow to
analyze current datasets of millions of cells across thousands of phenotypic dimensions. This proposal explores
the use of Bayesian hierarchical models to unlock robust and scalable analysis of single-cell CRISPR screen
data. We will design two tools based on this framework: one for single-cell CRISPR screens with gene expression
readouts (Aim 1.1) and one for optical pooled screens (Aim 2.1). Both tools will use state-of-the-art statistical
methodologies to quickly and robustly infer which perturbations affect which phenotypes. This task is a critical
step towards deepening our understanding of biological pathways and gene regulatory networks using these
large-scale perturbation datasets. We will design and benchmark our methods using public datasets from both
screen modalities to evaluate their performance and generalizability. To demonstrate our methods’ ability to
generate biological insights in clinically relevant contexts, we will apply our tools to new screens from our
collaborators to aid in nominating therapeutic targets for blood disorders (Aim 1.2) and neurological diseases
(Aim 2.2). The proposed work will be broadly useful to practitioners of single-cell CRISPR screens and will help
to democratize these complex screening approaches for widespread use.

## Key facts

- **NIH application ID:** 10891780
- **Project number:** 1F31HG013609-01
- **Recipient organization:** HARVARD MEDICAL SCHOOL
- **Principal Investigator:** Logan Blaine
- **Activity code:** F31 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $43,203
- **Award type:** 1
- **Project period:** 2024-07-01 → 2027-06-30

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10891780

## Citation

> US National Institutes of Health, RePORTER application 10891780, Robust and scalable Bayesian analysis tools for single cell CRISPR screens with sequencing- and imaging-based readouts (1F31HG013609-01). Retrieved via AI Analytics 2026-05-22 from https://api.ai-analytics.org/grant/nih/10891780. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*