Michael Hall
I am a Bioinformatics PhD student in Zam Iqbal’s lab at EMBL-EBI. I currently work on using nanopore data and genome graphs to better call variation in bacterial genomes and to compare pan-genomes.
Mycobacterium tuberculosis transmission cluster inference with Nanopore sequencing
Dataset
- Mtb WGS from culture
- Madagascar (118), South Africa (83), Birmingham (46)
- Matched Illumina and Nanopore (and 35 PacBio CCS)
- 150 after QC (7 PacBio)
Calibration of Nanopore variant filters
Aim: produce precision on par with Illumina
Calibration of Nanopore variant filters
Illumina SNPs called with COMPASS
Nanopore SNPs called with bcftools
Evaluate precision and recall with varifier for 7 PacBio samples
Different bcftools filters applied
Precision = fraction of calls that are correct
Recall = fraction of expected calls made correctly
Calibration of Nanopore variant filters
FP counts: 0, 0, 0, 1(0), 1(0), 1(0), 4(3)
Pairwise SNP distance
- Generate consensus sequences
- Ignore filtered, masked, missing, and null calls
- Pairwise distance matrix (one for each modality)
Aim: do we see a linear relationship?
Pairwise SNP distance
SNP threshold clustering
Aim: produce Nanopore clusters consistent with Illumina
- Thresholds: 0, 2, 5, and 12 (11 ONT)
- Connect two samples if SNP distance ≤ threshold
- Assess clustering based on 3 metrics
SNP threshold clustering
C_I =
A sample's Illumina cluster
C_N =
A sample's Nanopore cluster
Recall (SACR):
\frac{|C_I \cap C_N|}{|C_I|}
Precision (SACP):
\frac{|C_I \cap C_N|}{|C_N|}
SNP threshold clustering
Excess Clustering Rate (XCR):
\frac{|S_I - S_N|}{|S_I|}
S_I =
Set of Illumina singletons
S_N =
Set of Nanopore singletons
SNP threshold clustering
Nanopore doesn't miss any clustered samples
SACR: 1.0
SACP: 0.845
XCR: 0.031 (3/97)
SNP threshold clustering
SACR: 1.0
SACP: 1.0
XCR: 0.015 (2/137)
SACR: 1.0
SACP: 0.966
XCR: 0.008 (1/128)
SACR: 1.0
SACP: 0.949
XCR: 0.057 (7/122)
SACR: 1.0
SACP: 0.845
XCR: 0.031 (3/97)
Nanopore doesn't miss any clustered samples
Mixing technologies
Aim: can you cluster Illumina and Nanopore data?
Mixing technologies
Self-distance
Mixing technologies
Pairwise SNP distance
Mixing technologies
Simulate various mixture ratios
Nanopore-to-Illumina ratios: 0.01, 0.05, 0.1, 0.25, 0.5, 0.75, 0.9
Mixed SNP thresholds same as Illumina
For each ratio/threshold combo, do the following 1000 times:
- Randomly assign samples to a tech. in given ratio
- Calculate SACR, SACP, and XCR for given threshold
Simulate various mixture ratios
Simulate various mixture ratios
- SACR (median) 1.0 for all simulations
- Increasing Nanopore ratio moves median towards Nanopore-only SACP and XCR
- In the "worst" case, mixed tech. produces the same results as Nanopore-only
Summary
- Nanopore SNP precision on par with Illumina
- Nanopore SNP recall ~8% lower than Illumina
- No clustered samples are missed by Nanopore
- Mixing technologies is possible
Mtb transmission cluster inference with Nanopore sequencing
By Michael Hall
Mtb transmission cluster inference with Nanopore sequencing
Rolling results for our work checking concordance between Nanopore and Illumina for public health applications
