MorPhiC Policy on Genome Assembly

Approved by the Morphic Consortium Data Working Group: 2023-05-15

Approved by the Morphic Consortium Steering Committee: XXXX-XX-XX

The Morphic Consortium will provide uniformly processed data for production grade experiment types. It is important that data from different experiments and experiment types are processed with a consistent set of reference annotations to ensure integrability. This document describes the Human Genome Reference assembly selected by the Morphic Consortium DPC and DRACC, and outlines the motivation for this selection.

We propose to use the GRCh38 (Genome Reference Consortium Human Reference 38 Patch 15 (GCA_000001405.15) including the 25 assembled chromosomes (1-22, X, Y, M), the 127 unplaced contigs, and the 42 unlocalized contigs, but excluding the 261 alternative haplotypes. Since Epstein-Barr virus (EBV) very often turns up in human DNA sequencing and since this decoy assembly is included by the ENCODE consortium, we also propose to include the EBV assembly AC:AJ507799.2. This proposal will align our genome assembly exactly to the assembly used by ENCODE3 and ENCODE4.

This is a rapidly evolving area and this policy will be reviewed annually to assess transitioning to other reference assemblies, taking into consideration advances in analysis software as well as the cost of converting existing Morphic data to those alternatives.

Why GRCh38 rather than hg19?

• It is a more accurate assembly. In particular, ENCODE has found GRCh38 handles repeat sequences better, such that blacklists to filter repeats become less important.
• It is more future proof.

Why GRCh38 rather than T2T chm13?

• While T2T has several improvements over GRCh38 (e.g. overall higher quality scaffolds, fewer lower quality variants, fewer Mendelian violating variants, etc.), there is less mature software support at present and most external data resources will not shift across to T2T chm13 in the near future. Transition to T2T chm13 will be considered during each annual review.

Why not cell-line specific reference assemblies?

• This will be difficult for DPCs which have tens of cell lines, and the generation of high quality assemblies per cell line will be cost-prohibitive.

Why exclude alternative haplotypes?

• Standard aligners cannot utilize the extra information effectively. Reads mapping in these regions may be reported as non-unique alignment.
• We would like to use the same assembly as ENCODE.

Why not use hg19 or T2T in addition to GRCh38?

• Providing alternative mapping also will double our data processing costs (roughly $50k in fiscal year3 and going up in later years.)
• Providing alternative mapping also will require additional infrastructure development (UI tools to clearly denote which files are from which assembly) which will cost the DRACC software development team two months that can be used for other development.
• It will slow down the transition that the field is already going through.

How do we handle legacy data sets?

• Data from big consortia such as ENCODE are already available with respect to GRCh38.
• For data types for which we have implemented a uniform processing pipeline, the DRACC can process data from selected landmark publications to generate files in GRCh38. The DWG has asked all members of the consortium to nominate such papers for consideration by DWG process legacy data from other production omics experiment types (Repli-seq, ChIP-seq, ATAC-seq, DNAse-seq, RNA-seq, ChIAPET/PLAC-seq) once standards have been approved.