hg19 (GRCh37) vs. hg38 (GRCh38) Human Genome Reference Comparison Zuotian Tatum Department of Human Genetics Leiden University Medical Center
Timeline GRCh37: First release: Latest patch: GRCh38: First release: Feb 27, 2009 Latest patch: Jun 28, 2013 (p13) GRCh38: First release: Dec 24, 2013 Latest patch: Oct 14, 2014 (p1) http://www.ncbi.nlm.nih.gov/projects/genome/assembly/grc/human/data/
Content GRCh37.p13: Total bases: N50: Number of alternative loci: 3.23 Billion 2.99 Billion (without N) N50: 46 Million Number of alternative loci: 9 Non-nuclear genome: No GRCh38.p2: Total bases: 3.21 Billion 3.05 Billion (without N) N50: 67 Million Number of alternative loci : 261 Non-nuclear genome: Yes http://www.ncbi.nlm.nih.gov/projects/genome/assembly/grc/human/data/
UCSC tracks for GRCh38 UCSC RefSeq available since April 2014. Ensembl regulatory build available since September 2014. dbSNP 141 available since October 2014. ENCODE and FANTOM5 track hubs are still not available (Nov 2014).
New in GRCh38 release Three new sequence files, in addition to the standard assembly files: - GCA_000001405.15_GRCh38_top-level.fna.gz - GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.gz - GCA_000001405.15_GRCh38_full_analysis_set.fna.gz The analysis set files are created to avoid false mapping in NGS alignment pipelines.
GCA_000001405.15_GRCh38_top-level.fna.gz All the top-level objects in the full-assembly Chromosomes unlocalized scaffolds unplaced scaffolds alternate locus scaffolds mitochondrial genome The sequence identifiers are International Sequence Database Collaboration (INSDC) accession.versions and the definition lines are GenBank style. No sequences have been hard-masked.
GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.gz Chromosomes from the GRCh38 Primary Assembly unit. Note: the two PAR regions on chrY have been hard-masked with Ns. The chromosome Y sequence provided therefore has the same coordinates as the GenBank sequence but it is not identical to the GenBank sequence. Similarly, duplicate copies of centromeric arrays and WGS on chromosomes 5, 14, 19, 21 & 22 have been hard-masked with Ns. Mitochondrial genome from the GRCh38 non-nuclear assembly unit. Unlocalized scaffolds from the GRCh38 Primary Assembly unit. Unplaced scaffolds from the GRCh38 Primary Assembly unit. Epstein-Barr virus (EBV) sequence Note: The EBV sequence is not part of the genome assembly but is included in the analysis set as a sink for alignment of reads that are often present in sequencing samples.
GCA_000001405.15_GRCh38_full_analysis_set.fna.gz = GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.gz + alt-scaffolds from the GRCh38 ALT_REF_LOCI_* assembly units
Alt-loci add complexity to RNASeq quantification
Ideogram of GRCh38.p2
RNASeq quantification - Fragments (reads) per million per killobase (FPKM/RPKM) values to quantify gene expression - Unique mapping only Analysis tools do not distinguish allelic duplication from paralogous duplication - Non overlapping gene regions
To understand the effect of alt-loci on RNASeq quantification Compare alignment of chromosome 6 MHC region between - hg19 full set with 7 alt-loci - hg38 analysis set without alt-loci Sequence content are largely unchanged between hg19 and hg38.
Mapping/alignment for RNASeq hg19 hg38 mapped 14,655,299 14,704,427 mappedDiffChr 4,959 4,017 mappedPairProper 14,639,261 14,690,090 mappedPairProperPct 92.62 92.94 total 15,805,561 totalSplice 5,060,829 5,078,133 unmapped 1,150,262 1,101,134 hg19: with alt loci hg38: without alt loci
Effect of alt loci in RNASeq alignments Gene RPKM (hg38)
Distribution of RPKM difference
Major Histocompatibility complex region on chromosome 6
HLA-A hg19 full set – chr6 D1 hg19 full set – chr6_mann_hap4 D1 hg19 full set – chr6_qb1_hap6 D1 hg19 full set – chr6_dbb_hap3 D1
HLA-A hg19 full set – chr6 D1 D2 D3 hg38 analysis set D1 D2 D3
HLA-C hg19 full set D1 D2 D3 hg38 analysis set D1 D2 D3
HLA-DRA hg19 full set D1 D2 D3 hg38 analysis set D1 D2 D3
Major Histocompatibility complex region on chromosome 6 Class III
MHC Class III 700kb stretch, 60 genes. The most gene-dense region of the human genome > 14% coding ~ 72% transcribed Highly conserved Only a free have clearly defined and proven function
TNF hg19 full set – chr6 D1.control D1.treated hg38 analysis set – chr6 D1.control D1.treated
Highly variant immune regions retiled
LILRA3 moved to alt-loci in hg38 LILRB2 LILRA3 LILRA5 hg38 LILRB2 LILRA5
Phantom LILRA3
LILRA3 in hg19 LILRB5 Intergenic LILRB3 LILRA4
Gene length calculation We need gene length for calculating RPKM. If alignment uses alt loci RPKM would be artificially lowered for alt loci genes. If alignment does not alt loci Remove alt loci annotations from the official set.
Need more comprehensive approach to genome variation. Assembly model is neither haploid nor diploid Analysis tools penalize reads mapping to > 1 location do not distinguish allelic duplication from paralogous duplication A graph structure is a natural way to represent a population- based genome assembly
Conclusions RPKM values are highly correlated between hg19 and hg38. Analysis set is preferred for expression analysis. Additional analysis may be performed to use the alt-loci separately. Annotations for hg38 is still lacking and need contribution from the community. Improve modeling of genome variability in population.
Questions?