diff --git a/topics/single-cell/tutorials/scatac-standard-processing-snapatac2/tutorial.md b/topics/single-cell/tutorials/scatac-standard-processing-snapatac2/tutorial.md
index 8865a794d4e86e..82c21c4e927164 100644
--- a/topics/single-cell/tutorials/scatac-standard-processing-snapatac2/tutorial.md
+++ b/topics/single-cell/tutorials/scatac-standard-processing-snapatac2/tutorial.md
@@ -17,7 +17,7 @@ objectives:
 - Create a count-matrix from a 10X fragment file
 - Perform filtering, dimension reduction and clustering on AnnData matrices
 - Generate and filter a cell-by-gene matrix
-- Identify marker genes for the clusters and annotate clusters
+- Identify marker genes for the clusters and annotate the cell types
 time_estimation: 2H
 key_points:
 - Single-cell ATAC-seq can identify open chromatin sites
@@ -176,7 +176,7 @@ SnapATAC2 requires 3 input files for the standard pathway of processing:
 >  > 
 >  > 3. Rename the generated file to `Fragments 500 PBMC`
 >  > 4. Now you can continue with either the `fragments_file` from earlier or the new file `Fragments 500 PBMC`. 
->  >    - {% icon galaxy-info %} Please note that `Fragments 500 PBMC` only contains 500 {PBMC} and thus the clustering will produce different outputs compared to the outputs generated by `fragments_file` (with 5k PBMC). 
+>  >    - {% icon galaxy-info %} Please note that `Fragments 500 PBMC` only contains 500 {PBMC's} and thus the clustering will produce different outputs compared to the outputs generated by `fragments_file` (with 5k PBMC). 
 >  {: .hands_on}
 {: .details}
 
@@ -451,7 +451,7 @@ Doublets are removed by calling a customized [**scrublet**](https://github.com/A
 >    - The observed features of the "cells" are then compared to the simulated doublets and scored on their doublet probability. 
 >    - SnapATAC2's *pp.filter_doublets* then removes all cells with a doublet probability >50%. 
 >
-> ![Doublet removal with scrublet]({% link topics/single-cell/images/scatac-standard-snapatac2/doublets-and-scrublet.png %} "Scrublet simulates expected doublets and produces doublet scores for each cell.")
+> ![Doublet removal with scrublet]({% link topics/single-cell/images/scatac-standard-snapatac2/doublets-and-scrublet.png %} "Scrublet simulates expected doublets and produces doublet scores for each cell. ({% cite Wolock2019 %})")
 >
 {: .details}
 
@@ -502,13 +502,14 @@ Dimension reduction is a very important step during the analysis of single cell
 >
 > - Dimension reduction algorithms can be either linear or non-linear. 
 > - Linear methods are generally computationally efficient and well scalable. 
+>
 >   A popular linear dimension reduction algorithm is: 
 >     - **PCA** (Principle Component Analysis), implemented in **Scanpy** (please check out our [Scanpy]({% link topics/single-cell/tutorials/scrna-scanpy-pbmc3k/tutorial.md %}) tutorial for an explanation). 
 > - Nonlinear methods however are well suited for multimodal and complex datasets. 
 >     - in contrast to linear methods, which often preserve global structures, non-linear methods have a locality-preserving character. 
 >     - This makes non-linear methods relatively insensitive to outliers and noise, while emphasizing natural clusters in the data ({% cite Belkin2003%})
 >     - As such, they are implemented in many algorithms to visualize the data in 2 dimensions (f.ex. **UMAP** embedding).
-> - The nonlinear dimension reduction algorithm, through *spectral embedding*, used in **SnapATAC2** is a very fast and memory efficient non-linear algorithm ({% cite Zhang2024%}). 
+> - The nonlinear dimension reduction algorithm, through *matrix-free spectral embedding*, used in **SnapATAC2** is a very fast and memory efficient non-linear algorithm ({% cite Zhang2024%}). 
 >     - **Spectral embedding** utilizes an iterative algorithm to calculate the **spectrum** (*eigenvalues* and *eigenvectors*) of a matrix without computing the matrix itself. 
 {: .details}
 
@@ -524,7 +525,7 @@ The dimension reduction, produced by the algorithm *tl.spectral*, is required fo
 >
 >    > <comment-title> Distance metric </comment-title>
 >    >
->    > - The fast and well scalable *matrix-free spectral embedding* algorithm depends on the distance metric: `cosine`
+>    > - The fast and well scalable *"matrix-free spectral embedding"* algorithm depends on the distance metric: `cosine`
 >    {: .comment}
 >
 > 2. Rename the generated file to `Anndata 5k PBMC spectral` or add the tag {% icon galaxy-tags %} `spectral` to the dataset
@@ -968,7 +969,7 @@ To manually annotate the *Leiden* clusters, we will need to perform multiple ste
 
 
 # Conclusion
-{% icon congratulations %} Well done, you’ve made it to the end! You might want to consult your results with this [control history](https://usegalaxy.eu/u/timonschlegel/w/workflow---standard-processing-of-10x-single-cell-atac-seq-data-with-snapatac2), or check out the [full workflow](https://singlecell.usegalaxy.eu/u/timonschlegel/w/2combined-snapatac2) for this tutorial.
+{% icon congratulations %} Well done, you’ve made it to the end! You might want to consult your results with this [control history](https://singlecell.usegalaxy.eu/u/timonschlegel/h/test-of-5k-pbmc-tutorial-workflow), or check out the [full workflow](https://usegalaxy.eu/u/timonschlegel/w/workflow---standard-processing-of-10x-single-cell-atac-seq-data-with-snapatac2) for this tutorial.
 
 In this tutorial, we produced a count matrix of {scATAC-seq} reads in the `AnnData` format and performed: 
 1. Preprocessing: