enclone banner

enclone in cellranger

This page does two things:

  1. Show how to use the versions of enclone that were used by previous versions of Cell Ranger.
  2. Show the enclone changes that affect Cell Ranger output.

previous versions

Previous releases of enclone, matching particular Cell Ranger releases, may be found here:
5.0.0 (5.0.1 used the same enclone version)
6.0.0 (6.0.1 used the same enclone version).
7.0.0 (7.0.1 used the same enclone version).
To use these, click on the enclone_linux or enclone_macos link (on one of those pages) to download the binary executable file, then e.g. chmod +x enclone_macos to add execute permission, then run with e.g. enclone_macos.

enclone changes that affect cellranger

date git hash first Cell Ranger version change
Dec. 9, 2020 v0.5.23 f50783 6.0
  • Change the algorithm for deciding to use a donor reference allele, so that it checks all donor reference alleles for all V genes having the same name as the one originally assigned to a contig.
  • Change default MAX_DIFFS from 50 to 55. Change default MAX_DEGRADATION from 3 to 2. Change default MAX_SCORE from 1,000,000 to 500,000. Change default MAX_CDR3_DIFFS from 10 to 15.
  • Allow up to 5 (instead of 4) amino acid differences when joining chains. This results in less spurious chains in clonotypes.
Dec. 9, 2020 v0.5.24 1e1ef9 6.0
  • Add a doublet test, which removes some exact subclonotypes that appear to represent doublets.
  • Improve the disjoint orbit test.
Feb. 13, 2021 v0.5.38 a43c8c 6.1
  • To determine assignment of chains to columns in a clonotype table, we now rely primarily on the column information implied by the original joins of exact subclonotypes. This eliminates some artifactual behavior.
  • Change weak chain filtering to permit deletion of chains supported by up to 20 cells, rather than 5. This fixes some pathological cases involving clonotypes having many chains.
June 4, 2021 v0.5.66 635832 6.1 Tweak the order of default filters so as to increase the likelihood that clonotypes that are held together tenuously (or not at all) will be pulled apart. This has a tiny effect on our datasets (changing about 1 in 10^5 clonotypes), but could affect some user datasets disproportionally.
Sept. 20, 2021 v0.5.105 95c951 7.0 Add a new default filter called signature filtering that under certain rare circumstances will decompose complex clonotypes that are incorrectly glued together. This filter can dramatically affect certain datasets, but has almost no effect on typical data (less than one per million clonotypes tested). Please see enclone default filters.
Dec. 13, 2021 v0.5.135 803b05 7.0 Make three changes to clonotyping that increase its specificity:
  1. Reject joins where the number of differences in CDR3 is at least CDR3_MULT times the number of non-shared differences outside CDR3. The default value of CDR3_MULT is 5. This algorithmic change may be overridden by using a large value instead, e.g. by CDR3_MULT=100.
  2. Compute the number N in a different and simpler fashion, as 80^cd. The old behavior may be obtained by adding the argument OLD_MULT.
  3. Prevent merger in cases where light chain constant regions are different. The old behavior maybe be obtained by adding the argument OLD_LIGHT.
Jan. 5, 2022 v0.5.141 7e3d7b 7.0 Make five changes to BCR clonotyping that increase its sensitivity and specificity:
  1. Increase the precision of the calculation that computes the probability p that shared mutations occur by chance. This was returning the wrong answer in some cases. The old behavior is no longer available.
  2. Remove the cap on total differences between two exact subclonotypes in order for them to be joined. The old behavior may now be obtained by adding the argument MAX_DIFFS=55.
  3. Rather than impose a bound of 15 on the maximum number of nucleotide differences between CDR3 sequences, we now require at least 80% identity (out of the total CDR3 length). For both of these, the measurement first computes the total number of differences, taking into account one heavy and one light chain. The number 80 can be set to an alternate value using JOIN_CDR3_IDENT=.... The old behavior can be obtained by setting JOIN_CDR3_IDENT=0 and MAX_CDR3_DIFFS=15.
  4. Now compute the number N in a manner that takes into account CDR3 length. To do this, we first set cd1 to the number of heavy chain CDR3 nucleotide differences, and cd2 to the number of light chain CDR3 nucleotide differences. Let n1 be the nucleotide length of the heavy chain CDR3, and likewise n2 for the light chain. Then N is set to
    80^(42 * (cd1/n1 + cd2/n2)). The number 42 can be changed using the variable CDR3_NORMAL_LEN. The behavior preceding Dec. 13 is still available by via the argument OLD_MULT.
  5. The light chain constant region identity requirement is now relaxed to apply only when cd = 0. The relaxation cannot be turned off but the light chain requirement can still be turned off entirely using OLD_LIGHT.
Jan. 13, 2022 v0.5.142 b0adb3 7.0 Make three small changes to BCR clonotyping:
  1. Change the default value for MAX_SCORE to 100000.
  2. Fix bug that resulted in the constraint cd ≤ 15 still being imposed.
  3. Allow joins with high score to pass so long as they have at least 15 shared mutations. The parameter 15 is accessible as AUTO_SHARE.
Feb. 7, 2022 v0.5.147 34c334 7.0 Improve the correctness of V gene assignment. The changes comprise several technical changes to the alignment code and the addition of one gene IGHV3-9 to the VDJ reference. Also remove some gene copies from the reference that were identical except for extension on the 3' end.
Feb. 24, 2022 v0.5.151 37b7f0 7.0 Two more changes to the join algorithm.
1. If V gene names are different (after removing trailing *...), and either
• V gene reference sequences are different, after truncation on right to the same length
• or 5' UTR reference sequences are different, after truncation on left to the same length, then the join is rejected.
2. If the percent nucleotide identity on heavy chain FWR1 is at least 20 more than the percent nucleotide identity on heavy chain CDR1+CDR2 (combined), then the join is rejected.
3. Change the default for JOIN_CDR3_IDENT from 80 to 85.
Mar. 11, 2022 v0.5.156 e81193 7.1 Two more changes to the join algorithm.
1. A major change to the clonotype joining algorithm, which allows certain joins having high heavy chain CDR3 sharing. Please see enclone help how.
2. Delete the following genes from the human VDJ reference: IGHV4-30-2, IGKV1D-33, IGKV1D-37, IGKV1D-39, IGKV2D-28. These are identical to other genes in the reference, except that the reference provides a longer 5'-UTR in one case. We observe that clonotypes having one of these genes often have heterogeneous assignments between the gene and its duplicate, and that's bad, as it implies that assignment of genes to clonotypes in these cases is effectively random.
Mar. 12, 2024 8.0 From Cell Ranger 7.2 and onwards, Cell Ranger's dependency on Enclone has been removed. Cell Ranger now depends on enclone-ranger. This table is preserved for users working with older versions of Cell Ranger but will no longer be updated.