Expression and Methylation QC and PreProcessing RANDA STRINGER

  • Slides: 16
Download presentation
Expression and Methylation: QC and Pre-Processing RANDA STRINGER

Expression and Methylation: QC and Pre-Processing RANDA STRINGER

Epigenetics in Nutri. Gen • Examine epigenetic mediation of prenatal environment • Subset of

Epigenetics in Nutri. Gen • Examine epigenetic mediation of prenatal environment • Subset of Nutri. Gen cohorts • ~500 each from START and CHILD • Matched samples (where possible) • Integrate methylation and expression changes

Expression Data • Ongoing • Majority of samples available • START = 496 •

Expression Data • Ongoing • Majority of samples available • START = 496 • CHILD = 467 • QC/pre-processing is underway

Expression QC/Pre-Processing • Quality assessment • Probe boxplots (signal distributions, unusual samples) • Background

Expression QC/Pre-Processing • Quality assessment • Probe boxplots (signal distributions, unusual samples) • Background correction • Uses negative controls • Normalization • Batch effects • MDS plots/Com. Bat • Transformation • Probe filtering • Detection p < 0. 01 in > 50% of samples

Methylation Data • In analysis stage • QC/pre-processing complete* • *Pending further advancements in

Methylation Data • In analysis stage • QC/pre-processing complete* • *Pending further advancements in the field • Good final sample size for both • START = 506 • CHILD = 491

Methylation QC/Pre-Processing • Sample Quality • Compare reported vs. predicted sex • Remove samples

Methylation QC/Pre-Processing • Sample Quality • Compare reported vs. predicted sex • Remove samples where proportion of failed probes is > 0. 01 • Probe Quality • Remove probes that failed to be detected in > 5% of samples • Remove cross-hybridizing and polymorphic probes • Chen et al. 2013

Methylation QC/Pre-Processing • Normalization • 2 probe types with different distributions Infinium I Probe

Methylation QC/Pre-Processing • Normalization • 2 probe types with different distributions Infinium I Probe 2 different probes per Cp. G Infinium II Probe Single base extension at Cp. G Maksimovic et al. Genome Biology 2012

Type I Grn Probes Type II Probes

Type I Grn Probes Type II Probes

Methylation QC/Pre-Processing • Batch effects • Adjust for technical variation • Corrected by plate

Methylation QC/Pre-Processing • Batch effects • Adjust for technical variation • Corrected by plate • Cellular composition • Crucial issue in methylation studies • Cord blood not well characterized • Re. FACTor (Rahmani et al. , 2016) • Reference-free • Utilizes PCA

Other Considerations • Background correction • Bead count (> 3) • SNP probe definition

Other Considerations • Background correction • Bead count (> 3) • SNP probe definition • MAF > 0. 01 • Other normalization methods • BMIQ vs SWAN • Cellular composition adjustment

QC Summary Probes Samples START CHILD 512 511 Sex Check 5 14 Missingness 2

QC Summary Probes Samples START CHILD 512 511 Sex Check 5 14 Missingness 2 7 506 491 Initial Final START Initial Failed CHILD > 485 000 756 634 Polymorphic 70 889 Cross-Reactive 29 233 Final 393 400 393 449

Questions?

Questions?

Normalization Goal: reduce non-biological variation Equalizes probe intensity and signal distributions across arrays and

Normalization Goal: reduce non-biological variation Equalizes probe intensity and signal distributions across arrays and between colour channels New challenges with DNA methylation vs. gene expression techniques ◦ Systematic/technical variation ◦ Novel probe design Maksimovic et al. Genome Biology 2012

Cp. G Content Infinium II ≤ 3 Infinium I ≥ 3 Compressed β value

Cp. G Content Infinium II ≤ 3 Infinium I ≥ 3 Compressed β value distribution in Inf. II Solution: scale Infinium II probes to Inf. I probes Maksimovic et al. Genome Biology 2012

Subset Within-Array Normalization (SWAN) Allows Inf. I and Inf. II probes to be normalized

Subset Within-Array Normalization (SWAN) Allows Inf. I and Inf. II probes to be normalized together ◦ Subset of N Inf. I and Inf. II probes chosen based on underlying Cp. G content ◦ Separate methylated and unmethylated channels ◦ Mean intensity for each of 3 N calculated ◦ Inf. I and II probes adjusted separately by linear interpolation Maksimovic et al. Genome Biology 2012

Beta-MIxture Quantile normalization (BMIQ) Novel normalization method ◦ Fit 3 -state (U/H/M) to Inf.

Beta-MIxture Quantile normalization (BMIQ) Novel normalization method ◦ Fit 3 -state (U/H/M) to Inf. I and Inf. II probes separately ◦ Transform Inf. I U and M probes using the inverse of the cumulative beta distribution estimated from the respective Inf. II probes ◦ For H probes perform dilation transformation to fit the data into the gap Teschendorff et al. Bioinformatics 2012

Epigenetics DNA methylation II DNA methylation in plantsEpigenetics DNA methylation II DNA methylation in plants
The Methylation Landscape of Human Cancers Methylation patternsThe Methylation Landscape of Human Cancers Methylation patterns
Text Operations Preprocessing and Compression Introduction Document preprocessingText Operations Preprocessing and Compression Introduction Document preprocessing
Preprocessing Chapter 11 What is Preprocessing n PreliminaryPreprocessing Chapter 11 What is Preprocessing n Preliminary
Preprocessing Part 1 Preprocessing create a new dataPreprocessing Part 1 Preprocessing create a new data
Text Operations Preprocessing Introduction Document preprocessing to improveText Operations Preprocessing Introduction Document preprocessing to improve
Data Preprocessing Data Preprocessing p Why preprocess theData Preprocessing Data Preprocessing p Why preprocess the
Data Preprocessing 1 DWDM Data Preprocessing February 21Data Preprocessing 1 DWDM Data Preprocessing February 21
Preprocessing Part 1 Preprocessing create a new dataPreprocessing Part 1 Preprocessing create a new data
Text Preprocessing Steps in text preprocessing Obtaining textText Preprocessing Steps in text preprocessing Obtaining text
NLP Introduction to NLP Preprocessing Text Preprocessing RemovingNLP Introduction to NLP Preprocessing Text Preprocessing Removing
Data Preprocessing 1 The need of data preprocessingData Preprocessing 1 The need of data preprocessing
Data Preprocessing Agenda Why data preprocessing Data cleaningData Preprocessing Agenda Why data preprocessing Data cleaning
UNIT 1 Data Preprocessing Data Preprocessing Learning ObjectivesUNIT 1 Data Preprocessing Data Preprocessing Learning Objectives
Preprocessing Overview Preprocessing to enhance recognition performance inPreprocessing Overview Preprocessing to enhance recognition performance in