10x Genomics
Chromium Single Cell Gene Expression

Cell Ranger3.1, printed on 11/23/2024

Release Notes for Cell Ranger 3.1: Gene Expression & Feature Barcoding

Major Changes

• Feature Barcoding Only Analysis - It is now possible to run cellranger count using Cell Surface Protein (antibody captured) libraries without a GEX library. The previous version of Cell Ranger required a Gene Expression library along with a library generated by Feature Barcoding technology. However, the new version of Cell Ranger provides customers with flexibility to sequence either one of the libraries, or both. In particular, cell calling now works with antibody counts only, and all secondary analyses (PCA, t-SNE, UMAP, clusterings) work with antibody-only count matrix as well. More details are available on the Feature Barcoding Only Analysis page.

• UMAP based lower dimensionality projections of datasets analyzed by cellranger count are now produced in addition to the previously produced t-SNE projections. The projections are made available both as CSV files and as data that can be directly viewed in Loupe Browser. The parameters for the projection can also be modified and experimented with using cellranger reanalyze. This alternate visualization method has become increasingly popular for visualizing single cell data since the earliest report that used it. For more details, see the description in the algorithms overview section.

• New Web Summary Look - The Cell Ranger web_summary.html file has been updated to match the styles and formats of other 10x products. Compared to the old version users will notice new fonts and some aesthetic changes in the new version.

Minor Changes

• Bug Fix: If equal numbers of reads with given Barcode / UMI combination map to two genes, the assignment of the Barcode / UMI are now considered ambiguous and not reported in moleculeinfo.h5 or the count matrix. Previously they were reported _twice, once for each gene.
• Other minor bug fixes

Release Notes for Martian 3.2.3: Job Scheduling

• Fix a crash in cases where the mrp binary becomes unavailable on disk during a pipestance run.
• In addition to logging the type of filesystem for the pipestance directory, mrp will also log the type of filesystem for the martian bin directory (which is often different from the pipestance directory), and also the mount options for both directories.
• Regardless of --jobinterval setting, mrp will now never attempt to submit more than one job at a time to the queue in cluster mode.
• mrp will now shut down if the pipestance log file has been deleted, even if a new one has been created in its place. This prevents problems in the case where the pipestance directory (including the log and lock files) have been deleted.
• Memory cgroups limits are now detected, reported, and used as default limits where applicable. This should be especially helpful for users submitting mrp to a cluster such as SLURM which uses memory cgroups to prevent jobs from using too much memory, by preventing mrp from trying to use more than the job's allowance.
• Other small bug fixes and performance improvements.

Release Notes for Cell Ranger 3.1: V(D)J

Major Algorithmic Changes and Effects on Performance

• The assembly, annotation and cell calling algorithms have all been replaced, as have the reference sequences. However with noted exceptions, the interface is unchanged.

• Many changes were made to the assembly algorithm that allow it to achieve the same sensitivity using less data. After these changes, the recommended sequencing configuration was changed to 26 x 91 (from 150 x 150), while leaving the number of read pairs per cell fixed at 5000. This enables V(D)J, Gene Expression and Feature Barcoding libraries to be sequenced in a single run, thereby simplifying the workflow.

• The effect of the new changes varies considerably from sample to sample and we have added a discussion on Experimental Design that explains some of this. In some instances the number of productive pairs increases markedly if the same dataset is rerun with the new code.

• The old read configuration 150 x 150 is still supported and may be preferable for some users, because of pricing or availability, particularly for users who are running only V(D)J data. For 150 x 150, the recommended depth is proportionally lower, 2000 read pairs per cell.

• Many corrections were made to the Prebuilt reference sequences.

• Contig annotation has been improved in several ways. This includes more accurate detection of CDR3 regions, a more stringent full-length requirement, and a requirement that V segments begin with a start codon (coupled to reference sequence corrections). This could affect annotation for species other than human or mouse, having incomplete reference sequences.

• A productive pair is no longer declared in cases where there are three or more contigs having the same chain type (e.g. TRB, TRB, TRB). In such cases the GEM may contain two or more cells.

• Some new large clones are now reported, that were missed previously for a variety of reasons, including failure to align J segments having high somatic hypermutation.

• A productive pair is no longer declared in cases where three or more contigs share the same chain type (e.g. TRB, TRB, TRB). In such cases the GEM may contain two or more cells. In addition, certain clonal expansions of plasma cells are now contracted because the expansion represents mRNA leakage during processing, rather than a true biological expansion. Finally, requirements for small clones sharing a chain with a large clone have been tightened to reduce the likelihood of false clones arising from ambient mRNA or doublets. All of these changes correctly reduce the number of reported productive pairs (usually by a small fraction).

• Because of these changes, we recommend that customers rerun existing datasets using Cell Ranger 3.1 if possible.

• Because cell calling is changed, the denominator used for computing the Cells With Productive V-J Spanning Pair metric may have changed. For this reason, differences in performance between Cell Ranger 3.0 and 3.1 are better assessed using the Number of Cells With Productive V-J Spanning Pair metric.

• Cell Ranger 3.1 is significantly faster. There are five fewer stages in the pipeline.

Interface Changes

• Cell Count Confidence is no longer reported because we found that in some cases incorrect counts were reported with high confidence. Cell counting from V(D)J data alone is limited in accuracy because targeted cells having sufficiently low expression cannot be detected.

• Contigs Unannotated is no longer reported because all contigs are now annotated. The justification for this is that since enrichment uses primers binding to constant regions, bona fide contigs would be expected to have at least a C annotation.

• For species other than human or mouse, for which custom primers are needed, the sequences of the inner enrichment primers must now be supplied as a command-line argument.