Don-Isdale
commented
1 year ago 2023Sep07 - 2023Nov
Context and Discussion
Goals
- clarify what we mean by haplotype in Pretzel - how we define them, represent them, and so on.
- ability to relate genotyping platforms through haplotype
- a key use case will be for the user to be able select a haplotype in a dataset and see the relevant samples
- other use cases : ...
Most work on haplotypes will be considering a fixed set of positions in a single dataset - we need some higher level concepts as we want to talk about haplotypes independently of dataset (or at least have a concept that is independent).
i.e. compare "realised sets" of several datasets to see linkage, which suggests the broader / theoretic concept of haplotype, which is not visible directly in the given datasets but can be used to infer connection.
This section defines these terms :
- variant interval
- variant set
- realised haplotype ("haploallele"?)
A suggested model
- define a haplotype block or region via interval [start, end]. These could be called "variant intervals".
parent genome variant set (chr, pos, ref, alt, ) -> variant subset
- genotype dataset -> realised haplotypes
-
variant set may come from external definition, e.g. dbSNP, CropSNPdb, or from a genotype dataset,
-
a given genotype dataset will contain a subset of SNPs that variant set i.e. a genotype dataset under consideration may not contain all the SNPs in a variant set of the parent genome.
-
nulls : may be defined in alt of the variant set, or from a genotype dataset
-
for a given genotype dataset, a set of SNPs can be determined from the defined variant interval region [start, end]. i.e. the intersection of a variant interval and a genotype dataset => a set of SNPs which are a variant set.
For a given variant set and genotype dataset, the genotype values at the SNPs in the set are the realised haplotypes.
Representation in Database
The variant intervals can be represented as features in a dataset with the tag 'variantInterval'. Each feature has [start, end] which defines the variant interval.
variant set :
- dataset with tags 'SNP', 'variant'
- parent is the genome reference
- may not have all blocks / chromosomes of the parent
- feature :
-
- position
-
- values.{ref,alt}
Operations
Define a variant set by either :
- from an external source, e.g. upload via a spreadsheet or access from web api (CropSNPdb), or
- for a selected variant interval and a SNP or genotype dataset.
For a selected variant set :
- display realised haplotypes of the dataset
-
- This would be applied to the datasets which the user is viewing.
- select a realised haplotype
For a given variant set and genotype dataset, to calculate the realised haplotypes : Search may utilise the variant interval from which the variant set was defined, as an initial filter which leverages the feature position index.
For a selected realised haplotype :
-
filter the samples with the selected realised haplotypes, or
-
sort the sample columns according to distance from the selected realised haplotype
-
request the "realised haplotype" for any genotype file, which may simply return missing data - but the point is, the thing returned by that operation can be merged with what is returned from another file. This ensures that "realised haplotypes" for a given variant set can be merged together easily.
Comparison Operations
- for a variant interval / variant set, and some genotype datasets, report the realised haplotypes for the datasets.
-
- Further : request samples of each different realised haplotypes, i.e. 1 of each realised haplotype.
- for selected realised haplotype/s :
-
- request furthest sample from a realised haplotype.
-
- use selected realised haplotypes to colour (similar to the way that ref/alt colour, but 3rd possibility where the 2 are equal).
Those comparison operations represent the medium-term goal of this branch; as a possible intermediate step towards that :
- select 2 samples as parents, colour relative to those. bi-parental colouring
GUI
Operations :
- user selects a variant interval (perhaps via display of variant intervals in axis tracks), or a region between 2 SNPs.
Display realised haplotypes in the data for the region in a phylogeny / clade diagram and/or a graph of pie charts showing realised haplotypes proportions in the samples, as in :
from : https://www.researchgate.net/figure/Phylogenetic-tree-a-and-network-b-showing-the-genetic-relationships-among-the_fig4_280975921
To align with the sample column headers in the Pretzel Genotype Table, this diagram would be rotated 90deg right.
This enables the user to select a realised haplotype or group of realised haplotypes - the table can be narrowed to the samples with the selected realised haplotypes. Sample columns which don't have that realised haplotype can be filtered out or sorted to the right in order of distance from the selected realised haplotype. Colours can be assigned to realised haplotypes and shown in the table as a Haplotype Heatmap
Variations on the clade analysis graph :
- Notches on each branch represent the number of mutations between haplotypes, Fig. 7: clade analysis
- Figure 2 - The design used for nested clade analysis of the haplotype network
Similar :
- Fig 2 : Bayesian phylogenetic reconstruction. The phylogenetic analysis
- Fig 17 : Haplotype network
- Fig. 2 : Phylogeny and phylogeography
- Fig 2 : Phylogeny, haplotype network and genetic structure analyses
- Fig3 : nested haplotype structure and NCA
Evolutionary dynamics of genome size and content during the adaptive radiation of Heliconiini butterflies Figure 2 Figure 6
Optional : show a heat-map of realised haplotypes proportions in the samples, as in : Heatmap-of-haplotypes This could go at the top of the table, between the sample names (column headings) and the SNP rows.
- a view of haplotype intervals might be considered : display as intervals in axis tracks similar to this but the existing Pretzel track layering algorithm will produce a better display than this (which is just all intervals sorted by start, split into chunks and layered by index)
A few other views which may have some ideas to include :
- Figure 7 : Genealogical relationships among haplotypes
- Figure 2. HapMap Genome Browser This view shows the proportions of haplotype values at some SNPs.
The above graph visualisations seem to present the most accessible, self-explanatory and informative view for presenting realised haplotypes to the user. Keeping this out of the table will make it simpler to implement. The d3 functionality is fairly simple - they are just trees, the same tree with different layout.
Comparison of realised haplotypes
A visualisation may assist in seeing relationships between realised haplotypes of multiple genotypic datasets for the same variant set / variant interval.
This might be similar to these alignments of species phylogeny trees :
from Fig. 1
Similar :
- Figure 2 and Figure 3 from Multiple historical processes obscure phylogenetic relationships in a taxonomically difficult group (Lobariaceae, Ascomycota)
Following on from #343
Summary / Overview
dataset intersection :
12f369de dataset intersection : omit -r when requestOptions.isecDatasetIds is given. 030dc6de default enableFeatureFilters to true; update when it changes. access isec status. 0cd9e673 dataset intersection : order multiple datasets by selected SNP b94d986d [/6H] sort sample columns of all datasets together 78e7b93b [/6H] dataset intersection : selected dataset and one other 479852b4 re-indent 1e510218 [/3H] gui input of number of intersecting datasets ecc9e95d [6H/6H] dataset intersection : update displayed data; gui input of dataset positionFilter : show input value and update
2023Nov23 :
81d772c3 [/3H] enable SNP / Feature filters without selecting Features tab c6116158 [/2H] Use the type of currently selected sampleFilters and update. 09efe007 [/2H] add call rate filter on SNPs