New Analytical Tool Enhances Single Cell Data Interpretation

[seqadmin]

Boston University Team Develops DecontPro for Cleaner Data Analysis.

Researchers at Boston University Chobanian & Avedisian School of Medicine and College of Arts and Sciences have introduced a novel statistical model, DecontPro, to improve the analysis of single-cell data. This development addresses a critical challenge in the field of cellular indexing of transcriptomes and epitopes (CITE-seq), an RNA sequencing-based method used for simultaneous quantification of cell surface protein and transcriptomic data within single cells.

The Challenge of Background Noise in CITE-seq
CITE-seq's ability to concurrently study cell surface proteins and transcriptomic data offers valuable insights into cell types and disease states. However, a major limitation of this method is the significant levels of background noise which can obscure analysis. Background noise in CITE-seq can arise from various sources, often complicating the interpretation of data.

DecontPro: A Solution to Noise Interference
In response to this issue, the Boston University team, led by Joshua Campbell, Ph.D., an associate professor of medicine, developed DecontPro. Campbell describes DecontPro as a "statistical model that decontaminates two sources of contamination observed empirically in CITE-seq data." This model is designed to function as a quality assessment tool, assisting researchers in downstream analysis and enhancing understanding of the molecular causes of diseases.

How DecontPro Works
DecontPro operates by identifying and eliminating unwanted background noise from different sources. This includes noise from "spongelets", a novel type of artifact discovered by the researchers in their examination of several publicly available CITE-seq datasets. Spongelets contribute significantly to background noise across various datasets. DecontPro is capable of estimating and removing background noise arising from spongelets, ambient material in cell suspension, or non-specific binding of antibodies.

Masanao Yajima, Ph.D., a professor of the practice in the department of mathematics and statistics, elaborates on the model's design, stating, "DecontPro is a Bayesian hierarchical model. We carefully constructed it so that it can tease apart the signals from noise in single-cell datasets without being overly aggressive."

Implications for Future Research
The introduction of DecontPro promises to refine the process of single-cell data analysis significantly. By efficiently reducing the interference of background noise, this tool enables a clearer and more accurate interpretation of CITE-seq data. This advancement is expected to bolster research efforts in understanding cell behavior and disease mechanisms at a molecular level.

Read the full details of DecontPro in the journal Nucleic Acids Research.