Optimal tuning of weighted kNN- and diffusion-based methods for denoising single cell genomics data

Abstract

The analysis of single-cell genomics data presents several statistical challenges, and extensive efforts have been made to produce methods for the analysis of this data that impute missing values, address sampling issues and quantify and correct for noise. In spite of such efforts, no consensus on best practices has been established and all current approaches vary substantially based on the available data and empirical tests. The k-Nearest Neighbor Graph (kNN-G) is often used to infer the identities of, and relationships between, cells and is the basis of many widely used dimensionality-reduction and projection methods. The kNN-G has also been the basis for imputation methods using, e.g., neighbor averaging and graph diffusion. However, due to the lack of an agreed-upon optimal objective function for choosing hyperparameters, these methods tend to oversmooth data, thereby resulting in a loss of information with regard to cell identity and the specific gene-to-gene patterns underlying regulatory mechanisms. In this paper, we investigate the tuning of kNN-and diffusion-based denoising methods with a novel non-stochastic method for optimally preserving biologically relevant informative variance in single-cell data. The framework, Denoising Expression data with a Weighted Affinity Kernel and Self-Supervision (DEWÄKSS), uses a self-supervised technique to tune its parameters. We demonstrate that denoising with optimal parameters selected by our objective function (i) is robust to preprocessing methods using data from established benchmarks, (ii) disentangles cellular identity and maintains robust clusters over dimension-reduction methods, (iii) maintains variance along several expression dimensions, unlike previous heuristic-based methods that tend to oversmooth data variance, and (iv) rarely involves diffusion but rather uses a fixed weighted kNN graph for denoising. Together, these findings provide a new understanding of kNN-and diffusion-based denoising methods and serve as a foundation for future research.

A Machine-Learning Approach to Measure the Anterior Cruciate Ligament Injury Risk in Female Basketball Players

“…The selection of the right model parameters, such as kernel function for SVM, distance equation for kNN and number of splits for DT, appears to be a fundamental step for the classification process, since it can lead to a worsening of the algorithm performance. This result is in line with previous studies, in which it was demonstrated that the tuning of model parameters should be always conducted to avoid misclassifications [ 53 , 54 ]. After the right selection of the model parameters, by combining the results we can assess that l-SVM, q-SVM, f-kNN, m-DT and cx-DT meet all the criteria to be considered as optimum classifiers.…”

GABA-receptive microglia selectively sculpt developing inhibitory circuits

“…The Cre-Het control and GABA B1 R KO datasets were generated separately and, as such, exhibited stronger batch effects than the wild-type control and GABA B1 R KO datasets. For integrating the Cre-Het control and GABA B1 R KO datasets, DEWA ¨KSS was used (Tja ¨rnberg et al, 2020). Following standard pre-processing (normalization and log1p transformation), the optimal amount of principal components and k for the k-nearest neighbor graph was computed in DEWA ¨KSS over the BBKNN algorithm to integrate the conditions (Pola nski et al, 2020).…”

scBasset: Sequence-based modeling of single cell ATAC-seq using convolutional neural networks

“…Interestingly, we observed that even though SCALE correlation exceeded that of scBasset for baseline accessibility/expression, scBasset significantly outperforms SCALE when evaluated by differential accessibility/expression (p<7.25e-05). We hypothesize that SCALE’s reliance on cell-cell covariance encourages cells to be more similar to each other than they actually are and over-smooths (Tjarnberg et al, 2021; Ashuach et al, 2021). scBasset will be less prone to over-smoothing since each peak is considered only through its sequence.…”