Abstract
The count table, a numeric matrix of genes × cells, is the basic input data structure in the analysis of singlecell RNAsequencing data. A common preprocessing step is to adjust the counts for variable sampling efficiency and to transform them so that the variance is similar across the dynamic range. These steps are intended to make subsequent application of generic statistical methods more palatable. Here, we describe four transformation approaches based on the delta method, model residuals, inferred latent expression state and factor analysis. We compare their strengths and weaknesses and find that the latter three have appealing theoretical properties; however, in benchmarks using simulated and realworld data, it turns out that a rather simple approach, namely, the logarithm with a pseudocount followed by principalcomponent analysis, performs as well or better than the more sophisticated alternatives. This result highlights limitations of current theoretical analysis as assessed by bottomline performance benchmarks.
Similar content being viewed by others
Main
Singlecell RNAsequencing (RNAseq) count tables are heteroskedastic. In particular, counts for highly expressed genes vary more than for lowly expressed genes. Accordingly, a change in a gene’s counts from 0 to 100 between different cells is more relevant than, say, a change from 1,000 to 1,100. Analyzing heteroskedastic data is challenging because standard statistical methods typically perform best for data with uniform variance.
One approach to handle such heteroskedasticity is to explicitly model the sampling distributions. For data derived from unique molecular identifiers (UMIs), a theoretically and empirically wellsupported model is the gammaPoisson distribution (also referred to as the negative binomial distribution)^{1,2,3}, but parameter inference can be fiddly and computationally expensive^{4,5}. An alternative choice is to use variancestabilizing transformations as a preprocessing step and subsequently use the many existing statistical methods that implicitly or explicitly assume uniform variance for best performance^{3,6}.
Variancestabilizing transformations based on the delta method^{7} promise an easy fix for heteroskedasticity if the variance predominantly depends on the mean. Instead of working with the raw counts Y, we apply a nonlinear function g(Y) designed to make the variances (and possibly, higher moments) more similar across the dynamic range of the data^{8}. The gammaPoisson distribution with mean μ and overdispersion α implies a quadratic mean–variance relationship \({\mathbb{V}}{{{\rm{ar}}}}[Y]=\mu +\alpha {\mu }^{2}\). Here, the Poisson distribution is the special case where α = 0 and α can be considered a measure of additional variation on top of the Poisson. Given such a mean–variance relationship, applying the delta method produces the variancestabilizing transformation
Supplementary Information A1 shows the derivation. Practitioners often use a more familiar functional form, the shifted logarithm
This approximates equation (1), in particular if the pseudocount is y_{0} = 1 / (4α) (Supplementary Information A2).
An additional requirement is posed by experimental variations in sampling efficiency and different cell sizes^{9}, which manifest themselves in varying total numbers of UMIs per cell. Commonly, a socalled size factor s is determined for each cell and the counts are divided by it before applying the variancestabilizing transformation: for example, \(\log (y/s+{y}_{0})\)^{6,10,11}. There is a variety of approaches to estimate size factors from the data. Conventionally, they are scaled to be close to 1 (for example, by dividing them by their mean), such that the range of the adjusted counts is about the same as that of the raw counts. The simplest estimate of the size factor for cell c is
where the numerator is the total number of UMIs for cell c, g indexes the genes and \(L={\left({{{\rm{no}}.\, {\rm{cells}}}}\right)}^{1}{\sum }_{gc}{y}_{gc}\) is the average across all cells of these numerators.
Sometimes, a fixed value is used instead for L. For instance, Seurat uses L = 10,000, others^{12} have used L = 10^{6} calling the resulting values y_{gc}/s_{c} counts per million (CPM). Even though the choice of L may seem arbitrary, it matters greatly. For example, for typical dropletbased singlecell data with sequencing depth of ∑_{g}y_{gc} ≈ 5,000, using L = 10^{6} and then transforming to \(\log ({y}_{gc}/{s}_{c}+1)\) is equivalent to setting the pseudocount to y_{0} = 0.005 in equation (2). This amounts to assuming an overdispersion of α = 50, based on the relation between pseudocount and overdispersion explained in Supplementary Information A2. That is two orders of magnitude larger than the overdispersions seen in typical singlecell datasets. In contrast, using the same calculation, Seurat’s L = 10,000 implies a pseudocount of y_{0} = 0.5 and an overdispersion of α = 0.5, which is closer to overdispersions observed in real data. Yet, choosing L or y_{0} is unintuitive. Instead, we recommend parameterizing the shifted logarithm transformation in terms of the typical overdispersion, using the relation y_{0} = 1 / (4α) motivated above.
Hafemeister and Satija^{13} suggested a different approach to variance stabilization based on Pearson residuals
where \({\hat{\mu }}_{gc}\) and \({\hat{\alpha }}_{g}\) come from fitting a gammaPoisson generalized linear model (GLM),
Here, s_{c} is again the size factor for cell c, and β_{g,intercept} and β_{g,slope} are intercept and slope parameters for gene g. Note that the denominator in equation (4) is the s.d. of a gammaPoisson random variable with parameters \({\hat{\mu }}_{gc}\) and \({\hat{\alpha }}_{g}\).
A third set of transformations infers the parameters of a postulated generative model, aiming to estimate socalled latent gene expression values based on the observed counts. A prominent instance of this approach is Sanity, a fully Bayesian model for gene expression^{14}. It infers latent gene expression using a method that resembles a variational meanfield approximation for a lognormal Poisson mixture model. Sanity comes in two flavors: Sanity Distance calculates the mean and s.d. of the posterior distribution of the logarithmic gene expression; based on these, it calculates all cellbycell distances, from which it can find the knearest neighbors (kNN) of each cell. Sanity MAP (maximum a posteriori) ignores the inferred uncertainty and returns the maximum of the posterior as the transformed value. A related tool is Dino, which fits mixtures of gammaPoisson distributions and returns random samples from the posterior^{15}. Normalisr is a tool primarily designed for frequentist hypothesis testing^{16}, but as it infers logarithmic latent gene expression, it might also serve as a generic preprocessing method. Normalisr returns the minimum mean square error estimate for each count assuming a binomial generative model.
In this work, we analyze transformations for preprocessing UMIbased singlecell RNAseq data based on each of these approaches. We will first contrast the conceptual differences between them. In a second part, we benchmark the empirical performance of all approaches and provide guidelines for practitioners to choose among the methods. In the benchmarks, we also include a fourth preprocessing approach that is not transformationbased and directly produces a lowdimensional latent space representation of the cells: factor analysis for count data based on the (gamma)Poisson sampling distribution. An early instance of this approach, called GLM PCA, was presented by Townes^{4} and applied to biological data by Townes et al.^{17}. Recently, Agostinis et al.^{18} presented an optimized implementation called NewWave.
Results
There are multiple formats for each of the four approaches:

Among the delta methodbased variancestabilizing transformations, we considered the \({{{\rm{acosh}}}}\) transformation equation (1), the shifted logarithm equation (2) with pseudocount y_{0} = 1 or y_{0} = 1 / (4α) and the shifted logarithm with CPM. In addition, we tested the shifted log transformation with highly variable gene selection (HVG), z scoring (Z) and rescaling the output as suggested by Booeshaghi et al.^{19}.

Among the residualsbased variancestabilizing transformations, we considered the clipped and unclipped Pearson residuals (implemented by sctransform and transformGamPoi) and randomized quantile residuals. In addition, we tested the clipped Pearson residuals with HVG selection, z scoring and an analytical approximation to the Pearson residuals suggested by Lause et al.^{20}.

Among the latent gene expressionbased transformations (Lat Expr), we considered Sanity Distance and Sanity MAP, Dino and Normalisr.

Among the countbased factor analysis models (Count), we considered GLM PCA and NewWave.
Last, we include two methods as negative (Neg) controls in our benchmarks, for which we expect poor performance: the raw untransformed counts (y) and the raw counts scaled by the size factor (y / s).
Conceptual differences
A known problem for variancestabilizing transformations based on the delta method derives from the size factors. Figure 1a shows the first two principal components of a homogeneous solution of droplets encapsulating aliquots from the same RNA^{21} for representative instances of the delta method, residuals and latent expressionbased transformation approaches. Extended Data Fig. 1 shows the results for all transformations. Despite the size factor scaling, after the delta methodbased transformation, the size factor remained a strong variance component in the data (Extended Data Fig. 1b). In contrast, the other transformations better mixed droplets with different size factors. Intuitively, the trouble for the delta methodbased transformation stems from the fact that the division of the raw counts by the size factors scales large counts from droplets with large size factors and small counts from droplets with small size factors to the same value. This violates the assumption of a common mean–variance relationship. In Supplementary Information A3, we dissect this phenomenon more formally.
One of the motivations stated by Hafemeister and Satija^{13} for the Pearson residualsbased variancestabilizing transformation is that the delta methodbased transformations fail to stabilize the variance of lowly expressed genes. Warton^{22} provided a theoretical explanation for this fact. Indeed, Fig. 1b shows that the variance after transformation with a delta methodbased variancestabilizing transformation was practically zero for genes with a mean expression of <0.1. In contrast, after residualsbased transformation, the variance showed a weaker dependence on mean expression, except for very lowly expressed genes whose variance is limited by the clipping step (compare Pearson and Pearson (no clip) in Extended Data Fig. 2). The results of the latent expressionbased transformations were diverse, reflecting that these methods are not directly concerned with stabilizing the variance. Individual patterns ranged from higher variance for lowly expressed genes (Sanity Distance and Normalisr) to the opposite trend for Dino (Extended Data Fig. 2).
A peculiarity of the Pearson residuals is their behavior if a gene’s expression strongly differs between cell subpopulations. Figure 1c shows a bimodal expression pattern of Sftpc, a marker for type II pneumocytes. Unlike the transformations based on the delta method or latent expression models, the Pearson residuals are an affinelinear transformation per gene (equation (4)) and thus cannot shrink the variance of the highexpression subpopulation more than that of the lowexpression subpopulation (compare the Pearson residuals with y / s in Extended Data Fig. 3). This can affect visualizations of such genes and, in principle, other analysis tasks such as detection of marker genes or clustering and classification of cells.
An alternative is to combine the idea of delta methodbased variancestabilizing transformations with the generalized linear model residuals approach by using nonlinear residuals. We considered randomized quantile residuals^{23} (Extended Data Fig. 4 shows how they are constructed). Like Pearson residuals, randomized quantile residuals stabilized the variance for small counts (Extended Data Fig. 2), but in addition, they also stabilized the withingroup variance if a gene’s expression strongly differed across cells (Extended Data Fig. 3).
Such conceptual differences of the transformation approaches are important to understand when applying them to new data types or when developing new transformations; but for most practitioners, empirical performance will be of primary interest. We look at this in the next section.
Benchmarks
There is no contextfree measure of success for a preprocessing method, as it is contingent on the objectives of the subsequent analysis. For instance, if interest lies in identification of cell typespecific marker genes, one could assess the shape of distributions, such as in Fig. 1c, or the performance of a supervised classification method. Here, we considered the objective that arguably has been the main driver of singlecell RNAseq development and applications so far: understanding the variety of cell types and states in terms of a lowerdimensional mathematical structure, such as a planar embedding, a clustering, trajectories, branches or combinations thereof. For all of these, one can consider the knearest neighbor (kNN) graph as a fundamental data structure that encodes essential information. The next challenge is then the definition of ‘ground truth’. We designed our benchmarks upon reviewing previous benchmarking approaches. For instance, Breda et al.^{14} and Lause et al.^{20} employed synthetic or semisynthetic data. This is operationally attractive, but it is difficult to be certain about biological relevance. Hafemeister and Satija^{13} and Lause et al.^{20} used qualitative inspection of nonlinear dimension reduction plots. This can be informative, but is difficult to scale up and make objective. Germain et al.^{24} compared how well the transformations recovered a priori assigned populations, defined either through FACS or by mixing different cell lines. This is conceptually clean, but restricts analysis to a limited range of datasets that also may only offer a caricature view of cell diversity.
For all our benchmarks, we applied the transformations to the raw counts of each dataset listed below, computed a lowerdimensional representation of the cells using principalcomponent analysis (PCA), identified the kNNs of each cell as measured by Euclidean distance and, finally, computed the overlap of the thus obtained kNN graph with a reference kNN graph (Methods). We performed these three benchmarks:

Consistency. We downloaded ten 10x datasets from the Gene Expression Omnibus (GEO) database. As there was no formal ground truth, we measured the consistency of the results (a necessary, although not sufficient, condition for their goodness) by splitting the genes of each dataset into two disjoint subsets.

Simulation. We used four different previously published simulation frameworks and one adapted by us to generate a diverse collection of datasets for which we had full access to the true kNN graph.

Downsampling. We used five deeply sequenced datasets based on mcSCRB and Smartseq3, which we downsampled to sequencing depths typical for the 10x technology. We postulated that a proxy for ground truth could be constructed from the kNN graph inferred from the deeply sequenced data intersected across all transformations which we call reliable nearest neighbors. To our knowledge, this work presents the first instance of such an approach.
Extended Data Fig. 5 and the Supplementary Information give an overview of the datasets.
We tested 22 transformations—where applicable with an overdispersion fixed to 0, 0.05 and a genespecific estimate from the data—across four to eight settings for the number of dimensions of the PCA and measured the overlap with k = 10, 50 and 100 nearest neighbors. In total, we collected more than 61,000 data points. In addition to the results highlighted in the following, we provide an interactive website with all results for all tested parameter combinations.
Figure 2 shows the aggregated results for the three benchmarks for k = 50. Similar results were obtained for k = 10 and k = 100, shown in Extended Data Fig. 6.
In the consistency benchmark, the delta methodbased transformations performed better than the other transformations (Fig. 2a).
On the simulated data, the differences between the transformations looked less pronounced in Fig. 2b than for the other two benchmarks; however, this is a result of the aggregated view. For each particular simulation framework, large differences between the transformations appeared, but the results varied from simulation to simulation framework (Extended Data Fig. 7b) and averaged out in the aggregated view.
The results of the downsampling benchmark (Fig. 2c) agreed well with the trends observed in the simulation and the consistency benchmark. This benchmark is particularly informative because the data had realistic latent structures that were reliably detectable through the high sequencing depth. The downsampling produced data that resembles the more common 10x data in many characteristics: for example, UMIs per cell, proportion of zeros in the data and mean–variance relationship (Supplementary Table 1 and Supplementary Fig. 1). The main difference was that the suitable (high sequencing depth per cell) datasets we could access mostly consisted of only a few hundred cells, except for the 4,298cell shortinterfering RNA KD dataset (Extended Data Fig. 5).
The results in Fig. 2 are on a relative scale, which hides the magnitude of the differences. In Extended Data Fig. 7, we show the underlying results for each dataset on an absolute scale. The range of kNN overlaps was dataset dependent, ranging from 34 of 50 for the best performing transformation versus 9 of 50 for the negative control for the SUM149PT cell line consistency benchmark, to 2.9 of 50 versus 1.5 of 50 for the HEK downsampling benchmark. For the latter, the overall small overlaps were due to small sets of reliable nearest neighbors (Extended Data Fig. 8a,b). We also ran a version of the downsampling benchmark that only used the top two transformations per approach (Extended Data Fig. 8c,d), which increased the number of reliable nearest neighbors and confirmed the trends we saw in the full version.
In addition to the kNN overlap with the ground truth, we also calculated the adjusted Rand index (ARI) and the adjusted mutual information (AMI) for the five simulation frameworks. Extended Data Fig. 9a,b shows the aggregated results, which were similar to the results for the kNN overlap (Fig. 2b). Extended Data Fig. 9c,d show that the ARI and AMI had a larger dynamic range than the kNN overlap for datasets with a small number of distinct clusters; however, for datasets with a complex latent structure, the kNN overlap was more informative, which may reflect limitations of ARI and AMI to assess the recovery of gradual changes typical for many biological tissues.
The Random Walk simulation reproduced the benchmark based on which Breda et al.^{14} argued that Sanity was the best method for identifying the kNN of a cell (Fig. 5a of their paper). We found that the delta methodbased and residualsbased variancestabilizing transformations performed as well in this benchmark if we projected the cells to a lowerdimensional representation before constructing the kNN graph. In fact, Fig. 2d shows for four example datasets that the number of dimensions for the PCA was an important determinant of performance. This is because the dimension reduction acts as a smoothener, whose smoothing effect needs to be strong enough to average out uncorrelated noise (small enough target space dimension), but flexible enough to maintain interesting variation (large enough target space dimension).
The latent expressionbased transformations (except Normalisr) and the countbased factor analysis models were computationally more expensive than the delta method and residualsbased transformations. Figure 3a shows the CPU times for calculating the transformation and finding the kNN on the 10x human helper Tcell dataset with 10,064 cells. Sanity Distance took particularly long because its distance calculation, which takes into account the uncertainty for the nearest neighbor search, scaled quadratically with the number of cells (Fig. 3b). Across all benchmarks, the computations took 24 years of CPU time, of which the latent expressionbased transformations accounted for over 90%. The delta methodbased transformations were the fastest, especially if the overdispersion was not estimated from the data. The residualsbased transformations took somewhat more time, except for the analytic approximation of the Pearson residuals, which could be calculated almost as fast as the shifted logarithm. In terms of memory consumption, the delta methodbased transformations were most attractive because they retained the sparsity of the data.
In terms of uncovering the latent structure of the datasets, none of the other transformations consistently outperformed the shifted logarithm (Fig. 4a), one of the simplest and oldest approaches. In fact, when followed by PCA dimension reduction to a suitable target dimension, the shifted logarithm performed better than the more complex latent expressionbased transformations across all three benchmarks.
We found no evidence that additional postprocessing steps (rescaling the output of the shifted logarithm, selecting HVGs or equalizing the variance of all genes using z scoring) improved the results for identifying nearest neighbors (Fig. 4b). Lause et al.^{20} and Choudhary and Satija^{25} debated on how to best choose the overdispersion parameter. We found empirically that, for Pearson residuals and the \({{{\rm{acosh}}}}\) transformation, it is beneficial to have α > 0, but saw no clear benefits from estimating this parameter from the input data versus using a generic, fixed value such as 0.05 (Fig. 4c).
Last, we found that with increasing sequencing depth per cell, all methods generally had a better kNN overlap with the ground truth (Fig. 4d). This makes intuitive sense; with higher sequencing depth, the relative size of the sampling noise is reduced. Based on Fig. 1a, we might assume that delta methodbased transformations would perform particularly poorly at identifying the neighbors of cells with extreme sequencing depths; yet on three datasets, the shifted logarithm did not perform worse than other transformations for cells with particularly large or small size factors (Fig. 4d). We also considered the performance of the transformations as a function of cluster size (Extended Data Figs. 10); while we saw some interesting variation, we did not find that a single transformation performed consistently better or worse for small clusters.
Discussion
We compared 22 transformations, conceptually grouped into four basic approaches, for their ability to recover latent structure among the cells. We found that one of the simplest approaches, the shifted logarithm transformation \(\log (y/s+{y}_{0})\) with y_{0} = 1 followed by PCA, performed surprisingly well. We presented theoretical arguments for using the related \({{{\rm{acosh}}}}\) transformation or an adaptive pseudocount y_{0} = 1 / (4α), but our benchmarks showed limited performance benefits for these.
We recommend against using CPM as input for the shifted logarithm. We pointed out that for typical datasets, this amounts to assuming an unrealistically large overdispersion and in our benchmarks this approach performed poorly compared to applying the shifted logarithm to size factorscaled counts. We also advise against scaling the results of the shifted logarithm by the sum of the transformed values per cell as, for example, suggested by Booeshaghi et al.^{19}. In our hands (Extended Data Fig. 1), this additional operation failed to remove the confounding effect of the sequencing depth (the authors’ stated motivation for it) and did not improve the kNN recall performance.
The Pearson residualsbased transformation has attractive theoretical properties and, in our benchmarks, performed similarly well as the shifted logarithm transformation. It stabilizes the variance across all genes and is less sensitive to variations of the size factor (Extended Data Fig. 1b). The analytic approximation suggested by Lause et al.^{20} is appealing because it worked as well as the exact Pearson residuals but could be calculated faster. However, as seen in equation (4), the Pearson residualsbased transformation is affine linear when considered as a function per gene and this may be unsatisfactory for genes with a large dynamic range across cells. As an alternative, we considered randomized quantile residuals as a nonlinear transformation, but found no performance improvement. This result exemplifies that choosing a transformation for conceptual reasons does not necessarily translate into better downstream analysis results.
The use of the inferred latent expression state as a transformation and countbased latent factor models are appealing because of their biological interpretability and mathematical common sense. In particular, Sanity Distance is appealing because it does not have any tunable parameters; however, all these transformations performed worse than the shifted logarithm with a reasonable range of PCA dimensions in our benchmarks and some of the transformations were exceptionally computationally expensive (for example, the median CPU time of Sanity Distance was 4,500times longer than for the shifted logarithm).
Our results partially agree and disagree with previous studies. Germain et al.^{24} benchmarked many steps of a typical singlecell RNAseq analysis pipeline, including a comparison of clustering results obtained after different transformations against a priori assigned populations. In line with our findings, they reported that dimension reduction was of great importance. They went on to recommended sctransform (Pearson residuals) based on its good performance on the Zhengmix4eq dataset, which is a mixture of peripheral blood mononuclear cells sorted by surface markers using flow cytometry; however, it is not clear how generalizable this result is and our benchmarks do not support such a singling out of that method. Lause et al.^{20} considered the related Zhengmix8eq dataset, into which they implanted a synthetic rare cell type by copying 50 B cells and appending ten genes exclusively expressed in the synthetic cell type. They used kNN classification accuracy of the cell type averaged per cell type (macro F1 score; Fig. 5c of their paper) and averaged over all cells (online version of Fig. 5c). They found a performance benefit for the Pearson residuals over the shifted logarithm with the macro F1 score, but similar performance with regard to overall accuracy. The macro F1 score emphasizes the performance difference for the synthetic cell type, which seems somewhat construed and might not be a good model for most biologically relevant cell type and state differences. Instead of comparing clustering results to discrete cell type assignments, we have focused on the inference of the kNN of each cell, with the expectation that this enables consideration of more subtle latent structures than wellseparated, discrete cell types.
Pearson residuals and delta methodbased transformations weight genes differently; for example, Pearson residuals put more weight on lowly expressed genes than the delta method (Fig. 1b). This can lead to different downstream results, but our benchmarks did not indicate that any particular weighting is generally better; only that the delta methodbased transformation produced more consistent results on the 10x datasets.
We did not evaluate the impact of alternative size factor estimators. We also did not consider how suitable a transformation is for marker gene selection, because we are not aware of a suitable metric to determine success, as the utility of a marker gene hinges on its biological interpretability. For a recent effort to compare different marker gene selection methods, see Pullin and McCarthy^{26}.
Considerable research effort has been invested in the area of preprocessing methods for singlecell RNAseq data. To our surprise, the shifted logarithm still performs among the best. Our bottomline performance benchmark highlights current limitations of theoretical analysis of preprocessing methods, but also the utility of lowerdimensional embeddings of the transformed count matrix to reduce noise and increase fidelity. Interesting open questions include choosing among the many possible embedding methods and number of latent dimensions.
Methods
The delta method
The delta method is a way to find the s.d. of a transformed random variable.
If we apply a differentiable function g to a random variable X with mean μ, the s.d. of the transformed random variable g(X) can be approximated by
where \(a= {g}^{{\prime} }(\mu )\) is the slope of g at μ.
Now consider a set of random variables X_{1}, X_{2}, … whose variances and means are related through some function v, that is, \({\mathbb{V}}{{{\rm{ar}}}}[{X}_{i}]=v(\,{\mu }_{i})\), or equivalently \({\mathbb{S}}{{{\rm{d}}}}[{X}_{i}]=\sqrt{v(\,{\mu }_{i})}\). Then we can find a variancestabilizing transformation g by requiring constant s.d., \({\mathbb{S}}{{{\rm{d}}}}[g({X}_{i})]={{{\rm{const.}}}}\), which using the above approximation becomes
and can be solved by integration.
Transformations
We compared 22 transformations that can be grouped into four approaches.
The delta methodbased transformations were: the shifted logarithm (\(\log (y/s+1)\)); the acosh transformation (\({{\mathrm{acosh}}}\,(2\alpha y/s+1)\)); the shifted logarithm with pseudocount dependent on the overdispersion (\(\log (\,y/s+1/(4\alpha ))\)); the shifted logarithm with CPM (\(\log ({{{\rm{CPM}}}}+1)\)); the shifted logarithm with subsequent size normalization as suggested by Booeshaghi et al.^{19} (x_{gc}/u_{c}, where \({x}_{gc}=\log ({y}_{gc}/{s}_{c}+1)\) and u_{c} = ∑_{g}x_{gc}); the shifted logarithm with subsequent HVG selection (\(\log (y/s+1)\to {{{\rm{HVG}}}}\)); the shifted logarithm with subsequent z scoring per gene (\(\log (y/s+1)\to {{{\rm{Z}}}}\)); and the shifted logarithm with subsequent highly variable gene selection and z scoring per gene (\(\log (y/s+1)\to {{{\rm{HVG}}}}\to {{{\rm{Z}}}}\)). For all composite transformations, we first calculated the variancestabilizing transformation, then chose the HVGs and used the results without recalculating the variancestabilizing transformation.
To retain the sparsity of the output also if the pseudocount y_{0} ≠ 1, transformGamPoi uses the relation
Subtracting the constant \(\log {y}_{0}\) from this expression does not affect its variancestabilizing properties, but has the desirable effect that data points with y = 0 are mapped to 0.
The residualsbased transformations were: Pearson residuals implemented with the transformGamPoi package where each residual is clipped to be within \(\pm \sqrt{{{{\rm{no}}.\,{\rm{cells}}}}}\), as suggested by Hafemeister and Satija^{13} (Pearson); Pearson residuals with clipping and additional heuristics implemented by sctransform v.2, an analytic approximation to the Pearson residuals with clipping suggested by Lause et al.^{20} (Analytic Pearson); randomized quantile residuals implemented by transformGamPoi (Random Quantile); Pearson residuals without clipping implemented by transformGamPoi (Pearson (no clip)); Pearson residuals with clipping and subsequent HVG selection (Pearson → HVG); Pearson residuals with clipping and subsequent z scoring per gene (Pearson → Z); and Pearson residuals with clipping and subsequent HVG selection and z scoring per gene (Pearson → HVG → Z). For each composite Pearson residual transformation (with HVG and/or z scoring), we used the transformGamPoi implementation.
The latent expressionbased transformations were: Sanity with point estimates for the latent expression (Sanity MAP) and with calculation of all cellbycell distances taking into account uncertainty provided by the posteriors (Sanity Distance); Dino as provided in the corresponding R package; and Normalisr with variance normalization, implemented in Python, which we called from R using the reticulate package.
The countbased factor analysis models were: GLM PCA using the Poisson model and the gammaPoisson model with α = 0.05. In the figures, we show the results for the Poisson model unless otherwise indicated. We used the avagrad optimizer. We ran NewWave with 100 genes for the minibatch overdispersion estimation.
For the delta methodbased transformations and the residualsbased transformations calculated with the transformGamPoi package, we calculated the size factor s using equation (3).
We defined HVGs as the 1,000 most variable genes based on the variance of the transformed data.
For z scoring, we took the transformed values x_{gc} = g(y_{gc}) and computed \({z}_{gc}=\frac{{x}_{gc}{{\mathrm{mean}}}\,({{{{\boldsymbol{x}}}}}_{g})}{\sqrt{{{\mathrm{var}}}\,({{{{\boldsymbol{x}}}}}_{g})}}\), where mean and variance are the empirical mean and variance taken across cells.
In the overview figures (Figs. 2–4), we use a genespecific overdispersion estimate for all residualsbased transformations and for the delta methodbased transformations, which can handle a custom overdispersion; for GLM PCA, we use α = 0, because these settings worked best for the respective transformations. The latent expressionbased transformations and NewWave do not support custom overdispersion settings.
Conceptual differences
For the visualization of the residual structure after adjusting for the varying size factors, we chose a control dataset of a homogeneous RNA solution encapsulated in droplets^{21}. We filtered out RNAs that were all zero and plotted the first two principal components. Where applicable, we used genespecific overdispersion estimates. For visualizing the results of Sanity Distance, instead of the PCA, we used multidimensional scaling of the cellbycell distance matrix using R’s cmdscale function. We calculated the canonical correlation using R’s cancor function on the size factors and the first ten dimensions from PCA and multidimensional scaling.
The plots of the mean–variance relationship are based on the 10x human hematopoietic cell dataset^{27}. Where applicable, we used the genespecific overdispersion estimates. The panel of Sanity Distance shows the variance of samples drawn from a normal distribution using the inferred mean and s.d.
For the mouse lung dataset^{28}, we filtered out cells with extreme size factors (0.1s_{median} < s_{c} < 10s_{median}, where s_{median} is the median size factor). We also removed cells that did not pass the scran quality control criterion regarding the fraction of reads assigned to mitochondrial genes. To account for the fact that some transformations share information across genes, we applied all transformations to the 100 most highly expressed genes and three genes (Sftpc, Scgb1a1 and Ear2) known to be differentially expressed in some cell types according to the assignment from the original publication.
Benchmarks
The benchmarks were executed using a custom work scheduler for slurm written in R on CentOS7 and R 4.1.2 with Bioconductor v.3.14. The set of R packages used in the benchmark with exact version information was stored using the renv package and is available from the GitHub repository.
kNN identification and dimensionality reduction
To calculate the PCA, we used the irlba package. To infer the kNN, we used annoy, which implements an approximate nearest neighbor search algorithm. To calculate tdistributed stochastic neighbor embeddings (tSNEs), which we only used for visualization, we used the Rtsne package on data normalized with the shifted logarithm with a pseudocount of 1.
Consistency benchmark
We downloaded ten singlecell datasets listed in GEO browser after searching for the term mtx on 14 October 2021. All datasets are listed in the Data Availability section. To measure the consistency of the transformations, we randomly assigned each gene to one of two groups and processed the two resulting data subsets separately. We calculated the consistency as the mean overlap of the kNN for all cells.
Simulation benchmark
We used five frameworks to simulate singlecell counts in R: we ran dyngen^{29} using a consecutive bifurcating mode and the default parameters otherwise. We ran muscat^{30} with four clusters, a default of 30% differentially expressed genes with an average log fold change of 2 and a decreasing relative fraction of log fold changes per cluster. We ran scDesign2 (ref. ^{31}) with the 10x human hematopoietic cell dataset as the reference input with a copula model and a gammaPoisson marginal distribution. We simulated the Random Walk by translating the MATLAB code of Breda et al.^{14} to R and using the data by Baron et al.^{32} as a reference. For the Linear Walk, we adapted the Random Walk simulation and, instead of following a Random Walk for each branch, we interpolated the cells linearly between a random start and end point. For both benchmarks, we used a small nonzero overdispersion of α = 0.01 to mimic real data.
With each simulation framework, we knew which cells were the kNNs to each other. We calculated the overlap as the mean overlap of this ground truth with the inferred nearest neighbors on the simulated counts for all cells. Furthermore, we calculated the ARI and AMI by clustering the ground truth and the transformed values with the graphbased walktrap clustering algorithm from the igraph package.
Downsampling benchmark
We searched the literature for singlecell datasets with high sequencing depth and found five (one from mcSCRB, four from Smartseq3) that had a sequencing depth of more than 50,000 UMIs per cell on average. We defined reliable nearest neighbors as the set of kNNs of a cell that were identified with all 22 transformations on the deeply sequenced data (excluding the two negative controls). We used the downsampleMatrix function from the scuttle package to reduce the number of counts per cell to approximately 5,000, a typical value for 10x data. We considered only one setting for the overdispersion per transformation (instead of allowing multiple overdispersion settings for some transformations as in the other benchmarks). We ran all transformations that supported the setting, with a genespecific overdispersion estimate (except GLM PCA, which performed better with an overdispersion fixed to 0). Finally, we computed the mean overlap between the kNNs identified on the downsampled data with the set of reliable nearest neighbors for all cells with more than one reliable nearest neighbor.
kNN overlap
For all three benchmarks, we calculated overlaps between pairs of kNN graphs. Denoting their no. cell × no. cell adjacency matrices (a matrix of zeros and ones, where an entry is is one if a cell d is among the kNNs of cell c) by N^{1} and N^{2}, we defined their overlap as
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Code availability
An R package that implements the delta method and residualsbased variancestabilizing transformations is available at bioconductor.org/packages/transformGamPoi/. The code to reproduce the analysis and generate the figures is available at github.com/constae/transformGamPoiPaper and stored permanently with Zenodo^{45}. We provide an interactive website to explore the benchmark results at shinyportal.embl.de/shinyapps/app/08_singlecell_transformation_benchmark.
References
Grün, D., Kester, L. & Van Oudenaarden, A. Validation of noise models for singlecell transcriptomics. Nat. Methods 11, 637–640 (2014).
Svensson, V. Droplet scRNAseq is not zeroinflated. Nat. Biotechnol. 38, 147–150 (2020).
Kharchenko, P. V. The triumphs and limitations of computational methods for scRNAseq. Nat. Methods https://doi.org/10.1038/s4159202101171x (2021).
Townes, F. W. Generalized principal component analysis. Preprint at arXiv https://doi.org/10.48550/arXiv.1907.02647 (2019).
AhlmannEltze, C. & Huber, W. glmGamPoi: fitting gammaPoisson generalized linear models on single cell count data. Bioinformatics https://doi.org/10.1093/bioinformatics/btaa1009 (2020).
Amezquita, R. A. et al. Orchestrating singlecell analysis with Bioconductor. Nat. Methods 17, 137–145 (2020).
Dorfman, R.A. A note on the δmethod for finding variance formulae. Biometric Bull 1, 92 (1938).
Bartlett, M. S. The use of transformations. Biometrics 3, 39 (1947).
Lun, AaronT. L., Bach, K. & Marioni, J. C. Pooling across cells to normalize singlecell RNA sequencing data with many zero counts. Genome Biol. 17, 1–14 (2016).
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNAseq data with DESeq2. Genome Biol. 15, 550 (2014).
Borella, M., Martello, G., Risso, D. & Romualdi, C. PsiNorm: a scalable normalization for singlecell RNAseq data. Bioinformatics 38, 164–172 (2022).
Luecken, M. D. & Theis, F. J. Current best practices in singlecell RNAseq analysis: a tutorial. Mol. Syst. Biol. 15, e8746 (2019).
Hafemeister, C. & Satija, R. Normalization and variance stabilization of singlecell RNAseq data using regularized negative binomial regression. Genome Biol. 20, 1–15 (2019).
Breda, Jérémie, Zavolan, M. & van Nimwegen, E. Bayesian inference of gene expression states from singlecell RNAseq data. Nat. Biotechnol. https://doi.org/10.1038/s4158702100875x (2021).
Brown, J., Ni, Z., Mohanty, C., Bacher, R. & Kendziorski, C. Normalization by distributional resampling of high throughput singlecell RNAsequencing data. Bioinformatics 37, 4123–4128 (2021).
Wang, L. Singlecell normalization and association testing unifying CRISPR screen and gene coexpression analyses with Normalisr. Nat. Commun. 12, 1–13 (2021).
Townes, F. W., Hicks, S. C., Aryee, M. J. & Irizarry, R. A. Feature selection and dimension reduction for singlecell RNAseq based on a multinomial model. Genome Biol. 20, 1–16 (2019).
Agostinis, F., Romualdi, C., Sales, G. & Risso, D. NewWave: a scalable R/Bioconductor package for the dimensionality reduction and batch effect removal of singlecell RNAseq data. Bioinformatics 38, 2648–2650 (2022).
Booeshaghi, A. S., Hallgrímsdóttir, I. B., GálvezMerchán, A. & Pachter, L. Depth normalization for singlecell genomics count data. Preprint at bioRxiv https://doi.org/10.1101/2022.05.06.490859 (2022).
Lause, J., Berens, P. & Kobak, D. Analytic pearson residuals for normalization of singlecell RNAseq umi data. Genome Biol. 22, 1–20 (2021).
Svensson, V. et al. Power analysis of singlecell RNAsequencing experiments. Nat. Methods 14, 381–387 (2017).
Warton, D. I. Why you cannot transform your way out of trouble for small counts. Biometrics 74, 362–368 (2018).
Dunn, P. K. & Smyth, G. K. Randomized quantile residuals. J. Comput. Graph. Stat. 5, 236–244 (1996).
Germain, PierreLuc, Sonrel, A. & Robinson, M. D. pipeComp, a general framework for the evaluation of computational pipelines, reveals performant single cell RNAseq preprocessing tools. Genome Biol. 21, 1–28 (2020).
Choudhary, S. & Satija, R. Comparison and evaluation of statistical error models for scRNAseq. Genome Biol. 23, 1–20 (2022).
Pullin, J. M. & McCarthy, D. J. A comparison of marker gene selection methods for singlecell RNA sequencing data. Preprint at bioRxiv https://doi.org/10.1101/2022.05.09.490241 (2022).
Bulaeva, E. et al. MYCinduced human acute myeloid leukemia requires a continuing IL3/GMCSF costimulus. Blood 136, 2764–2773 (2020).
Angelidis, I. et al. An atlas of the aging lung mapped by single cell transcriptomics and deep tissue proteomics. Nat. Commun. 10, 1–17 (2019).
Cannoodt, R., Saelens, W., Deconinck, L. & Saeys, Y. Spearheading future omics analyses using dyngen, a multimodal simulator of single cells. Nat. Commun. 12, 1–9 (2021).
Crowell, H. L. et al. Muscat detects subpopulationspecific state transitions from multisample multicondition singlecell transcriptomics data. Nat. Commun. 11, 1–12 (2020).
Sun, T., Song, D., Li, WeiVivian & Li, JingyiJessica scDesign2: a transparent simulator that generates highfidelity singlecell gene expression count data with gene correlations captured. Genome Biol. 22, 1–37 (2021).
Baron, M. et al. A singlecell transcriptomic map of the human and mouse pancreas reveals interand intracell population structure. Cell Syst. 3, 346–360 (2016).
Kathiriya, J. J. et al. Human alveolar type 2 epithelium transdifferentiates into metaplastic KRT5^{+} basal cells. Nat. Cell Biol. 24, 10–23 (2022).
Nomaru, H. et al. Single cell multiomic analysis identifies a Tbx1dependent multilineage primed population in murine cardiopharyngeal mesoderm. Nat. Commun. 12, 1–19 (2021).
De Santis, R., Etoc, F., RosadoOlivieri, E. A. & Brivanlou, A. H. Selforganization of human dorsalventral forebrain structures by light induced SHH. Nat. Commun. 12, 1–11 (2021).
Pal, B. et al. Single cell transcriptome atlas of mouse mammary epithelial cells across development. Breast Cancer Res. 23, 1–19 (2021).
Porritt, R. A. et al. NLRP3 inflammasome mediates immunestromal interactions in vasculitis. Circ. Res. 129, e183–e200 (2021).
Panebianco, C. J., Dave, A., Charytonowicz, D., Sebra, R. & Iatridis, J. C. Singlecell RNAsequencing atlas of bovine caudal intervertebral discs: discovery of heterogeneous cell populations with distinct roles in homeostasis. FASEB J. 35, e21919 (2021).
Qian, Y. et al. ZEB1 promotes pathogenic Th1 and Th17 cell differentiation in multiple sclerosis. Cell Rep. 36, 109602 (2021).
Lu, X. et al. Identification of conserved SARSCoV2 spike epitopes that expand public cTfh clonotypes in mild COVID19 patients. J. Exp. Med. https://doi.org/10.1084/jem.20211327 (2021).
Bagnoli, J. W. et al. Sensitive and powerful singlecell RNA sequencing using mcSCRBseq. Nat. Commun. 9, 1–8 (2018).
HagemannJensen, M. et al. Singlecell RNA counting at allele and isoform resolution using Smartseq3. Nat. Biotechnol. 38, 708–714 (2020).
Larsson, AntonJ. M. et al. Transcriptional bursts explain autosomal random monoallelic expression and affect allelic imbalance. PLoS Comput. Biol. 17, e1008772 (2021).
Johnsson, P. et al. Transcriptional kinetics and molecular functions of long noncoding RNAs. Nat. Genet. 54, 306–317 (2022).
AhlmannEltze, C. Online supplement of comparison of transformations for singlecell RNAseq data. Zenodo https://doi.org/10.5281/zenodo.7504146 (2022).
Hotelling, H. Relations between two sets of variates. Biometrika 28, 321–377 (1936).
Acknowledgements
We thank S. Anders for extensive discussions about variancestabilizing transformations and how to benchmark preprocessing methods. We thank the three anonymous reviewers, E. van Nimwegen and D. Kobak, whose feedback on an earlier version helped to improve the paper. This work has been supported by the EMBL International PhD Program (C.A.E.), by the German Federal Ministry of Education and Research (CompLS project SIMONA under grant agreement no. 031L0263A) (C.A.E. and W.H.) and the European Research Council (Synergy Grant DECODE under grant agreement no. 810296) (C.A.E. and W.H.).
Funding
Open access funding provided by European Molecular Biology Laboratory (EMBL).
Author information
Authors and Affiliations
Contributions
C.A.E. and W.H. conceived the idea for the study and wrote the final paper. C.A.E. performed all computations with feedback from W.H.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Methods thanks Dmitry Kobak and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Lei Tang and Lin Tang, in collaboration with the Nature Methods team. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 Confounding effect of size factors on PCA embedding of a homogeneous dataset.
(A) Scatterplots of the first two principal components of the transformed data colored by the sequencing depth (expressed as a normalized size factor on a logarithmic scale) per cell. The data are from droplets that encapsulate a homogeneous RNA solution and thus the only variation is due to technical factors like sequencing depth^{21}. The annotation at the bottom of the plot shows the canonical correlation coefficient ρ^{46} between the size factor and the first ten principal components. A lower canonical correlation that the variancestabilizing transformation more successfully adjusts for the varying size factors; a canonical correlation of ρ = 1 means that the ordering of the cells along some direction in the first 10 PCs is entirely determined by the size factor. (B) Collection of the canonical correlations from the annotations of each panel in A displayed as a bar chart for easy visual comparison.
Extended Data Fig. 2 Meanvariance relations per gene for a 10x hematopoietic cell dataset.
Scatterplots of the variance per gene after applying the variancestabilizing transformation against the means of the 10x human hematopoietic cell dataset subset to 400 cells and 5000 genes. Note that the scale of the y axis differs for the raw counts, \(\log ({{{\rm{CPM}}}}+1)\), \(\log (y/s+1)/u\), Pearson (no clip), Sanity MAP, Dino and Normalisr for esthetic purposes. Points that exceed the y axis scale are drawn on the top of each facet.
Extended Data Fig. 3 Histograms of the transformed values for a gene with a bimodal expression pattern.
Counts from cells identified as type II pneumocytes are shown in purple and a matching number of counts from all other cell types are shown in gray.
Extended Data Fig. 4 Schematic representation of how randomized quantile residuals are constructed.
In the first step, a GammaPoisson distribution (black line) is fitted to the observed counts. Then, the quantiles of the GammaPoisson distribution are matched with the quantiles of a standard normal distribution by comparing their respective cumulative density functions (CDFs). This obtains a mapping from the raw count scale to a new, continuous scale. The two colored bars (orange for y = 2, yellow for y = 21) exemplify this mapping. The nonlinear nature of the CDFs ensures that small counts are mapped to a broader range than large counts. This helps to stabilize the variance on the residual scale. Furthermore, the randomization within the mapping sidesteps the discrete nature of the counts.
Extended Data Fig. 5 tSNE plots of each dataset used for the benchmarks.
The cells are colored by clustering using the walktrap clustering algorithm. For the consistency data we clustered the counts after transformation with the shifted logarithm. For the simulation data, we clustered the ground truth. For the downsampling data, we clustered the deeply sequenced data after transformation with the shifted logarithm.
Extended Data Fig. 6 Benchmark results for 10 and 100 nearest neighbors.
Plot of the aggregate results of the consistency (A, B), simulation (C, D) and downsampling benchmarks (E, F) for k = 10 and k = 100, respectively. The results for each dataset are broad to a common scale by normalizing to the mean k nearest neighbor overlap per dataset. The colored points show the averages across the datasets, each with 5 replicate random data splits (small, gray points).
Extended Data Fig. 7 The unaggregated benchmark results.
The unaggregated results from the consistency (A), simulation (B) and downsampling benchmarks (C) for k = 50. The gray points show the raw results from the five replicates per dataset; the colored points show their mean. The dashed vertical line indicates the mean kNN overlap per dataset and is the reference used to aggregated the results as shown in Fig. 2AC. The title of each facet indicates the number of dimensions used for the PCA per dataset, which we chose based on the complexity of the dataset.
Extended Data Fig. 8 Inference of the reliable nearest neighbors for the deeply sequenced datasets.
(A) Heatmaps of the average kNN overlap for all transformation pairs. (B) Histograms of the number of reliable neighbors per cell (that is, the neighbors among the 50 kNN that were identified by all 22 transformations). The dark shaded bars show the cells that were used to calculate the overlap with the downsampled version of the data in Extended Data Fig. 7C. (C) Histograms of the number of reliable neighbors per cell only considering the two two transformations per approach (that is, the neighbors among the 50 kNN that were identified by 8 transformations listed in (D)). (D) The unaggregated results for the downsampling benchmarks using the same settings as in Extended Data Fig. 7C.
Extended Data Fig. 9 Results of the simulation benchmark in terms of cluster similarity.
Plots of the results using the adjusted Rand index (A) and the adjusted mutual information (B) instead of the kNN overlap. (CD) Scatterplots facetted by simulation framework that compares the results for the kNN overlap with the adjusted Rand index and adjusted mutual information, respectively. Each point is one replicate for the transformation results of that dataset colored by the transformation approach. The black dashed line shows the linear fit and the number at the bottom of each plot is the correlation coefficient. The horizontal dashed line is the mean ARI / AMI that is used for forming the relative performance in (A) and (B). The vertical dashed line is the mean kNN overlap and corresponds to the vertical dashed line in Extended Data Fig. 7B.
Extended Data Fig. 10 kNN overlap as a function of cluster size.
The datasets were clustered using walktrap clustering. Extended Data Fig. 5 shows the cells colored by cluster assignment for all datasets. (A) kNN overlap of the two halves of the human neural progenitor dataset stratified by cluster. (B) kNN overlap with the ground truth for the scDesign2 simulation stratified by cluster. (C) kNN overlap with the deeply sequenced data for the siRNA knockdown dataset stratified by cluster.
Supplementary information
Supplementary Information
Supplementary Fig. 1, Supplementary Table 1 and mathematical details A1–A3.
Source data
Source Data Fig. 1
Excel file with three sheets for each panel containing the underlying data.
Source Data Fig. 2
Excel file with one sheet containing the overall benchmark results and one sheet with the data regarding the effect of the PCA dimension.
Source Data Fig. 3
Excel file with one sheet containing the CPU and elapsed time for the GSE179831 dataset and one sheet with raw data for the duration versus dataset size plot.
Source Data Fig. 4
Excel file with the data underlying the pairwise comparison.
Source Data Extended Data Fig. 1
Excel file with raw data.
Source Data Extended Data Fig. 2
Commaseparated file with raw data.
Source Data Extended Data Fig. 3
Commaseparated file with raw data.
Source Data Extended Data Fig. 5
Commaseparated file with raw data.
Source Data Extended Data Fig. 6
Commaseparated file with raw data.
Source Data Extended Data Fig. 7
Commaseparated file with raw data.
Source Data Extended Data Fig. 8
Excel file with raw data.
Source Data Extended Data Fig. 9
Commaseparated file with raw data.
Source Data Extended Data Fig. 10
Commaseparated file with raw data.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
AhlmannEltze, C., Huber, W. Comparison of transformations for singlecell RNAseq data. Nat Methods 20, 665–672 (2023). https://doi.org/10.1038/s41592023018141
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41592023018141
This article is cited by

The effect of data transformation on lowdimensional integration of singlecell RNAseq
BMC Bioinformatics (2024)

Systematic evaluation with practical guidelines for singlecell and spatially resolved transcriptomics data simulation under multiple scenarios
Genome Biology (2024)

Computational immunogenomic approaches to predict response to cancer immunotherapies
Nature Reviews Clinical Oncology (2024)

Prediction of tumorreactive T cell receptors from scRNAseq data for personalized T cell therapy
Nature Biotechnology (2024)

Representing and extracting knowledge from singlecell data
Biophysical Reviews (2024)