Assignment 7: Single-cell genomics. Bio /02/2018

Transcription

1 Assignment 7: Single-cell genomics Bio /02/2018

2 Assignment 7: Single-cell genomics Input Genotypes called from several exome-sequencing datasets derived from either bulk or small pools of cells (VCF format) Goals & output Write a VCF parser Analyze the variation in genotype coverage across samples Estimate the allelic dropout rate Filter genotypes for high quality variants 2

3 Protocol Input datasets Genotype calls from 9 exome-sequencing datasets generated from a single man Sample type Blood Live sperm Dead sperm Standard protocol 1 1 (1M sperm) 1 (1M sperm) MALBAC protocol - 3 biological replicates (each 200 sperm) 3 biological replicates (each 200 sperm) 3

4 Input file: sperm_genotype_calls.vcf Anatomy of a VCF file VCF contains more than alternative alleles, e.g., 1 & 2 VCF has been filtered for SNVs Autosomes (no sex or mitochondrial chromosomes) Adapted from Mayank s 2016 slides 4

5 Format of VCF records Name 1 CHROM Brief description The name of the sequence (typically a chromosome) on which the variation is being called. This sequence is usually known as 'the reference sequence', i.e. the sequence against which the given sample varies. 2 POS The 1-based position of the variation on the given sequence. 3 ID The identifier of the variation, e.g. a dbsnp rs identifier or just. if unknown. Multiple identifiers should be separated by semi-colons without white-space. 4 REF The reference base (or bases in the case of an InDel at the given position on the given reference sequence. 5 ALT The list of alternative alleles at this position. 6 QUAL A quality score associated with the inference of the given alleles. 7 FILTER A flag indicating which of a given set of filters the variation has passed. 8 INFO An extensible list of key-value pairs (fields) describing the variation. See below for some common fields. Multiple fields are separated by semicolons with optional values in the format: "<key>=[,data]". 9 FORMAT An (optional) extensible list of fields for describing the samples. See below for some common fields. + SAMPLEs For each (optional) sample described in the file, values are given for the fields listed in FORMAT 5

6 Why VCF? VCF is the worst variant call representation, except for all the others 6

7 Multiple Annealing and Looping Based Amplification Cycles (MALBAC) Purpose Performs linear whole genome amplification The original DNA is preferentially amplified Amplification bias is reduced Input Picograms of high molecular weight DNA (10-100kb) Specialized MALBAC primers: 35 nt long: 8 random and 27 common nts Bst polymerase (displacement activity) dntps Output Micrograms of amplified DNA Adapted from Dr. Conrad s lecture slides 7

8 MALBAC primers contain random nts so they can bind randomly to the template New DNA copy has complementary ends so it can loop and thus can no longer be a template MALBAC primers contain common nts so vanilla PCR can be used Zong, et al. Science Adapted from Dr. Conrad s lecture slides

9 Assignment TODOs Write a script called genotype_analysis.py to parse a VCF to do the following Calculate the mean and variance in genotype coverage for each dataset Plot the coverage distribution for each dataset 9

10 Boxplots (a.k.a. box-and-whisker plots) Useful for visualizing (multiple) distributions (median, spread of data, skewness, and outliers) Non-parametric Anatomy of a boxplot 10

11 How to create boxplots with matplotlib Output Matbplotlib boxplot tutorial 11

12 Assignment TODOs (cont.) Write a script called genotype_analysis.py to parse a VCF to do the following Calculate the mean and variance in genotype coverage for each dataset Plot the coverage distribution for each dataset OPTIONAL: use named command-line parameters 12

13 Getting fancy with usage statements: named command-line parameters Example usage with positional parameters $ bedtools intersect A.bed B.bed Example usage with named parameters $ bedtools getfasta -fi A.fa -bed B.bed -fo C.fa Why are named parameters useful? The order of the parameters doesn t matter Makes it easy to have (many) optional parameters Improves readability 13

14 How to use command-line named parameters Don t waste your time coding this yourself! Use the standard argparse module instead! Useful features of argparse Parses sys.argv so you don t have to Automatically generates help menus Automatically creates a --help option Automatically issues errors when invalid arguments are given. Examples: Options that are mutually exclusive Specific choices, e.g.,.csv or.tsv Specific variable types, e.g., an integer Short and long options can be used, e.g., --h vs. --help Sub-command options, e.g., bedtools intersect & bedtools getfasta Argparse tutorial 15

15 argparse example: the help message $ python3 example_argparse.py --help usage: example_argparse.py [-h] -a FIRST -b SECOND [-o {add,subtract}] Perform arithmetic optional arguments: -h, --help show this help message and exit -a FIRST, --first the first number -b SECOND, --second the second number -o {add,subtract}, --operator {add,subtract} Operation to perform on the numbers 16

16 argparse example: running scripts $ python3 example_argparse.py --first 1 --second 2 --operator add 3.0 $ python3 example_argparse.py -a 1 -b 2 -o add 3.0 $ python3 example_argparse.py -a STRING -b 2 -o add usage: example_argparse.py [-h] -a FIRST -b SECOND [-o {add,subtract}] example_argparse.py: error: argument -a/--first: invalid float value: 'STRING' 17

17 argparse example: the code 18

18 Assignment TODOs (cont.) Write a script called genotype_analysis.py to parse a VCF to do the following Calculate the mean and variance in genotype coverage for each dataset Plot the coverage distribution for each dataset OPTIONAL: use named command-line parameters Estimate the allelic dropout rate for each dataset 19

19 Allelic dropout (ADO) Loss of one or both alleles during amplification Decreases genotyping accuracy Hou, et al. 2012, Cell, report ADO rate of 11% with MDA + shotgun whole genome seq Same paper reports 43% using MDA + exome sequencing Solution could be filtering or explicit statistical model Adapted from Dr. Conrad s lecture slides

20 Allelic dropout (ADO) Random non-amplification of an allele TRUTH Homo: C/C C 1 C 1 C 1 C 2 C 2 C 2 Variant calling Homo: C/C C 1 C 2 C 1 C 1 C 1 Homo: C/C C 2 C 2 C 2 Homo: C/C

21 Allelic dropout (ADO) Random non-amplification of an allele TRUTH Het: A/C A C A C A C Variant calling Het: A/C A C A A A Homo: A/A C C C Homo: C/C

22 Calculating the ADO rate of a sample ADO occurs when 1) the reference is heterozygous AND 2) the sample is homozygous ADO rate = # of genotypes with ADO # of genotypes with ADO + # of genotypes w/o ADO # of genotypes with ADO TRUTH SAMPLE 0/0 0/0 0/1 0/1 1/1 1/1./../. Condition 1 Condition 2 # of genotypes w/o ADO TRUTH SAMPLE 0/0 0/0 0/1 0/1 1/1 1/1./../. Condition 1 Condition 2

23 Assignment TODOs (cont.) Write a script called genotype_analysis.py to parse a VCF to do the following Calculate the mean and variance in genotype coverage for each dataset Plot the coverage distribution for each dataset OPTIONAL: use named command-line parameters Estimate the allelic dropout rate for each dataset Filter genotypes for high quality variants (see the assignment for explicit instructions) 24

24 Transitions and transversions 25

25 Assignment 7: requirements Due Friday, 03/23/2018 at 10 AM Your submission directory should contain A README.txt with the answers to the questions and the commands/work you used to arrive at the answer A commented genotype_analysis.py A plot (variant_coverage_distributions.png) that is appropriately scaled with labelled axes and an informative title There is an extra credit option! Your code should be able to run on genomic.wustl.edu Do a final run in your submission directory to check 26