Detecting and comparing genomic compartments Cyril Kurylo , Sylvain - PowerPoint PPT Presentation

Detecting and comparing genomic compartments Cyril Kurylo , Sylvain Foissac , Matthias Zytnicki

Genomic structures Chromosome territories Segregation of untangled chromosomes A/B compartments Impact on gene expression Topologically Associating Domains Co-regulation domains Loops Interaction of regulatory elements Chromatin Compaction of DNA Accessibility to transcription DNA Do ğ an & Liu, 2018 Genetic information 1

Analysing compartmentalization Motivations Compartment changes linked to transcriptional changes Large White neonatal mortality due to poor fetal development 2

Using Hi-C to expose compartments Rao et al., 2014 3

Using Hi-C to expose compartments 4

Using Hi-C to expose compartments 4

Using Hi-C to expose compartments 4

Using Hi-C to expose compartments 4

Using Hi-C to expose compartments 4

Using Hi-C to expose compartments 5

Analysing compartmentalization Ambitions Computationally detect compartments Using replicates Providing a confidence measure Statistical comparison across conditions Data 2 conditions — 90 and 110 days of development 3 Hi-C replicates per condition 6

Hi-C DOC: Detection Of Compartments with replicates available at github.com/mzytnicki/HiCDOC Distance Cyclic Knight- Constrained Concordance Loess P-value Loess Ruiz k-means Measure Regression Matrix normalization Compartment detection Comparison 7

Correctly normalizing Hi-C matrices Technical biases Sequencing depth Restriction enzyme Cyclic Loess Difference Difference Cross-linking conditions multiHiCcompare Experiment quality Genomic distance in bins Genomic distance in bins Cyclic Knight- Distance Loess Ruiz Loess Matrix normalization Compartment detection Comparison 8

Correctly normalizing Hi-C matrices Biological biases GC content 8 1 Double stochastic Restriction site distribution 1 3 transformation Repeated sequences 1 6 Knight-Ruiz 9 1 5 1 8 3 6 9 5 1 1 1 1 1 Cyclic Knight- Distance Loess Ruiz Loess Matrix normalization Compartment detection Comparison 9

Correctly normalizing Hi-C matrices Distance effect Proximity between regions Loess regression Interaction Distance Cyclic Knight- Distance Loess Ruiz Loess Matrix normalization Compartment detection Comparison 10

Detecting compartments Constrained k-means K = 2 Predicted compartments B compartment A compartment Matrix normalization Compartment detection Comparison 11

Comparing compartmentalization between conditions B compartment Concordance at 90 days A compartment B compartment Concordance at 110 days A compartment Matrix normalization Compartment detection Comparison 12

Comparing compartmentalization between conditions P-value Distribution of the differences when the compartment doesn’t change Probability of observing a difference between concordances as extreme or more extreme when the compartment doesn’t change Differences between concordances 90 days 110 2.5% 2.5% days 0 Median differences for predicted compartment changes (with constrained k-means) Matrix normalization Compartment detection Comparison 13

Conclusion and perspectives Ambitions achieved Preliminary Results Perspectives Computationally detect compartments Predicted compartment Analyse genes changes in switching regions Using replicates Ongoing statistical analysis Publish method and results Providing a quantitative measure for our data Statistical comparison across conditions 14

Thank You github.com/mzytnicki/HiCDOC

Concordance comparison

Gene density 90 days 110 days Gene density (# genes / kb) Compartment

PCA detection Lieberman-Aiden et al., 2009