Single-cell RNA sequencing (scRNA-seq) allows researchers to analyze gene expression at the individual cell level, offering valuable insights into cellular diversity. However, one major challenge in scRNA-seq is dropout events, where some genes appear inactive simply because their expression wasn’t detected—not because they aren’t being used. This can lead to misleading results in studies of cell function and identity.

Researchers from the Wuhan Institute of Technology have developed AcImpute, a new method designed to improve the accuracy of scRNA-seq data analysis by addressing the common problem of oversmoothing in data imputation.

Data imputation refers to techniques used to fill in missing values in datasets. In scRNA-seq, missing values (dropouts) can distort clustering, classification, and gene expression studies. Traditional imputation methods smooth the data to estimate missing values, but they often oversmooth, blurring the differences between cells and reducing biological variability.

AcImpute overcomes this problem by enhancing constraints on smoothing weights, allowing for a more refined balance between restoring missing values and preserving genuine cellular differences.

The outline of AcImpute

Step 1: Normalize the matrix. Step 2: Select high-variant genes and proceed with principal component analysis (PCA) dimensionality reduction. Step 3: Apply KNN based strategy to data dimensionally reduced by PCA to obtain neighboring cells and calculate the inter-cellular transition probability matrix. Step 4: Obtain the power matrix by averaging the normalized matrix over its neighboring cells. Step 5: Obtain the final imputation matrix through the computation of formula X^ij=Norm(Mi.Pij) * (XN).j⁠.

The researchers compared AcImpute with nine other imputation methods and found that it:

• Restores gene expression more accurately, filling in missing values without introducing artificial biases.
• Preserves inter-cell variability, maintaining meaningful differences between individual cells.
• Prevents oversmoothing, ensuring that distinct cell types remain identifiable.
• Improves clustering and trajectory inference, making it easier to classify cells and track their developmental pathways.

By improving data quality in scRNA-seq, AcImpute helps scientists gain a more accurate picture of cellular functions, which is crucial for studying complex biological systems, including cancer progression, immune responses, and stem cell differentiation. This new approach enhances the reliability of computational analyses, leading to better biological insights and potential medical breakthroughs.

With AcImpute, researchers now have a more powerful tool for refining scRNA-seq data, leading to more reliable gene expression studies and better decision-making in biological and medical research. Future developments could further optimize imputation techniques, ensuring that scRNA-seq remains a gold standard for single-cell analysis.

Availability – The code is available at https://github.com/Liutto/AcImpute.