Eff fficient processing of Hi Hi-C data an and ap application to - PowerPoint PPT Presentation

Eff fficient processing of Hi Hi-C data an and ap application to can ancer Nicola las Serv rvant, PhD Institut Curie, INSERM U900, Mines ParisTech, PSL-Research University 04th of Decembre 2019 Hi-C days, Toulouse

2 Spa patial organization onof the he geno nome me How w are 2 metersof DNA packed into ntoa 10µm dia iameternuc ucle leus us ? Nucleosome (histone/DNA) H1 Histone Chromatin fibre Adapted from Annunziato et al. 2008

3 Different nt levels of spa patial organiza zation Nucleus

Hi Hi-C capt ptures the he chr hrom omatinconf nfor ormation on within the he nucleus Cells population Lieberman-Aiden et al. 2009, Rao et al. 2014 Contact chr1 chr4 chr2 chr3 chr5 chr6 … Frequencies chr6 High A B C Low A +1 Bin size A B +1 B C Whole genome map Intrachromosomalcontact maps at differentresolutions 4

5 Genom nome organization onand nd Hi Hi-C chr2 chr4 chr1 chr3 chr2 Lieberman-Aiden et al. 2009, Dixon et al. 2012, Rao et al. 2014

6 Genom nome organization onand nd Hi Hi-C ~100 Kb ~1 Mb Lieberman-Aiden et al. 2009, Dixon et al. 2012, Rao et al. 2014

7 Topol opolog ogical Associated doma omains ns (TAD ADs) The topological domains (TADs) have been described as the functional units of the genome organization, able to promote enhancer/promoter interactions. Contact Probability

8 'Hi-C'-basedexpe periments Method Main features References For mapping whole-genomechromatin interaction in a cell Lieberman-Aiden et al. (2009) Hi-C population; proximity ligation is carried out in a large volume Similar to Hi-C, except that proximity ligation is carried out Kalhor et al. (2011) TCC on a solid phase-immobilized proteins For mapping chromatin interactions at the single-cell level Nagano et al. (2013) Single-cell Hi-C Proximity ligation is carried out in the intact nucleus Rao et al. (2014) In situ Hi-C Combines 3C with a DNA capture technology ; equivalent to Hughes et al. (2014) Capture-C high-throughput4C Chromatin is fragmented with DnaseI; proximity ligation is Ma et al. (2015) Dnase Hi-C carried out on a solid gel Combine Dnaseor in situ DnaseHi-C with a capture Ma et al. (2015) Targeted Dnase Hi-C technology Chromatin is fragmented with micrococcalnuclease Hsiech et al. (2015) Micro-C Chromatin is fragmented with DnaseI; proximity logation is Deng et al. (2015) In situ DNAse Hi-C carried out in the intact nucleus Combines 3C with a DNA capture technology ; equivalent to Mifsud et al. (2015) Capture-Hi-C high-throughput5C Detecting genome-widechromatin interaction mediated by Mumbach et al. (2016) HiChIP a particular protein ; equivalent to ChAI-PET

9 Ready-to to-us use Hi Hi-C Kits https://www.qiagen.com https://arimagenomics.com/kit https://www.phasegenomics.com

Whi hichappr proachfor or whi hichpur urpo pose ?

Whi hichappr proachfor or whi hichpur urpo pose ? Captu pture Hi Hi-Cprotocol ( Franck et al. 2016 ) i.e. Hi-C library combined with capture of a dedicated genomic region

12 Que uestion ons ? 1. How to efficiently process Hi-C data? 2. Are there any specific computational challenges in analyzing Hi-C data from cancer samples ?

Wha hat doe oes Hi-C data look ook like ? Illumina paired-end sequencing Hi-C Fragments PE Sequencing

14 Cha hallenges in Hi Hi-Cdata proc ocessing IMR90 chr6 240 - Hi-C Fragments 0- IMR90 chr6 PE Sequencing Bin size = 500 Kb How to process Hi-C data in an easy and effic icie ient way taking into account ; - The huge amount of data - The evolutionof protocols - The computationalressources

15 Reads ds mapp pping ng strategy Hi-C Fragments PE Sequencing End-to-end genome Alignment Aligned Reads Unmapped Reads Trim 3’ RS site End-to-end genome Alignment Aligned Reads Unmapped Reads

16 Detectionof validinteraction on prod oduc ucts Hi-C Fragment Valid Pairs Invalid Pairs FR Singleton RF Dangling End FF RR Self Circle R=Reverse / F=Forward Dumped Pairs + Filtering on : • Insert size • Restriction fragment size • MAPQ • etc.

17 Bui uildi ding ng cont ntact maps ps bins There is currently no consensus about how to (efficiently) store the contact maps A Hi-C contact map is : • Usually very sparse bins • Symmetric We therefore propose to use a standard triplet sparse format to store only half of the non-zero contact values. Sparse Complete Dense (MB) Sparse Symmetric (MB) (MB) 1M 25 98 49 500Kb 77 363 182 150Kb 818 1 900 934 40Kb 12 000 3 800 1 900 20Kb 45 000 5 300 2 700 5Kb >100 000 ?? 8 600 4 300

18 Hi Hi-C formats .hic ic file iles (Juic uicer, Juic icebo box) .cool file les (coole ler, hig igla lass) • • Flexibility to store one or multiple matrices with Contact matrices in multiple resolutions and summary statistics stored in one file varying bin sizes • • python library Java and C bindings • • Command line tools Command line tools • • HDF5, which has native bindings in practically all Extant suite of analysis tools • Extant visualization tool. languages • out of memory iterative matrix balancing, that can work on very large matrices.

19 Hi Hi-C data nor ormalization All high-througthut techniques are subject to technical and experimentalbiases The iterativecorrection (ICE) method is a widely used approach for Hi-C data normalization. This method is based on the assumption that each locus should have the same probability of interaction genome-wide , and is in theory able to correct for any bias in the contact maps. 1 1 1 1

HiC-Pro – proc ocessing ng of Hi Hi-C/HiChiPdata Dedicated Detection of Valid Generates raw and Quality Reads Mapping Interaction normalized Controls Strategy Products contact maps • Easy-to-use Highlyusedin the last years • Optimized and scalable • Flexible Available at https://github.com/nservant/HiC-Pro • Support most protocols Forum and discussion at • Open to contribution https://groups.google.com/forum/#!forum/hic-pro • Compatible with many downstream analysis software

Bui uildi ding ng Efficient and ndRepr prod oduc ucibl ble Wor orkflows For facilities Highly optimised pipelines with excellent reporting. Validated releasesensure reproducibility.​ For users Portable, documentedand easy to use workflows. Pipelines that you can trust. For developers Companiontemplates and tools help to validateyour code and simplify common tasks.​

Analys ysis ispipe ipelin lines: Commun unit ity: 29 organisations over the world • Nextflow-based pipelines More than 90 contributors • High level of reproducibily • Strict Guidelines • 17 released pipelines • 19 under development

First version of nf-core Hi-C pipeline released ! V1.1.0 = Nextflow HiC-Pro version • Automatic installation • Natively support most schedulers • Natively compatible with conda, docker, singularity • Efficient tasks management • Reads can be automaticaly splitted by chuncks to speed the processing

Plans for the next xt versions : • TADs calling​ (which methods ?) • Compartment Calling​ • Detection of significant contacts • ​Specific pipelines for Hi -C based assembly ? Cancer Hi-C ? Contribution is welcome !​

25 Questions ? 1. How to efficiently process Hi-C data? 2. Are there any specific computational challenges in analyzing Hi-C data from cancer samples ?

Hi Hi-C on canc ncer data So far, most of the studies were dedicated to normal cell … and a few ones started to investigate chromatin structure of Breast and Prostate cancer using Hi-C

27 Alterations ns in canc ncer (epi pi)geno nomics

28 TADs are biol olog ogicallyrelevant • TADs disruption leads to new enhancer/promoter contacts • Abnormal enhancer/promoter contacts can have strong phenotypic impacts • Structural variants can disrupt TADs structure Luipanez et al. 2015, Franke et al. 2016

29 Organization on of canc ncer genom omes? Non-coding DNA mutation Structural Variants Valton & Dekker, 2016

30 Hi Hi-C, a good ood tool ol to stud udy CNVs ?

31 Cha hallenges in Hi-C canc ncer data? Hi-C Impacts? Simulation? Copy Number Normalization Variants Impacts?

Hi Hi-C C – Wha hat do we coun ount? chr6 HOX genes cluster i j Whole genome map Intrachromosomal contact Topological maps domains In the context of a diploid genome If i and j belong to the same chromosome C ij = 2 cis + 2 transH If i and j belong to different chromosomes C ij = 4 trans 32

Gene neralization n to pol olyploid genom omes N i = N j = 1 N i = N j = 2 N i = N j = 3 If chr i = chr j , C ij = 1 cis If chr i = chr j , C ij = 2 cis + 2 transH If chr i = chr j , C ij = 3 cis + 6 transH If chr i ≠ chr j , C ij = 1 trans If chr i ≠ chr j , C ij = 4 trans If chr i ≠ chr j , C ij = 9 trans N i = N j If chr i = chr j , C ij = N i cis + N i (N j -1) transH If chr i ≠ chr j , C ij = N i N j trans 33

Extens nsion on to o Canc ncer geno nome N = 4 If i and j belong to the same chromosomal segment C ij = N i cis + N i (N j -1) transH N = 2 34