Visualization and Interaction With Complex Multi-Scale Biological Data and How I Even Got to Work in Data Visualization Research David Kouřil Visualization II Invited Talk March 8, 2022 • My Path to Visualization Research • My PhD Projects • Current Work as Post-doc: Chromatin Visualization • Ad: Work With Us My Path to Visualization Research Background • Undergraduate (Be, Mgr.) at Fl MU • Erasmus in Vienna, VisGroup • PhD at TU Wien (March 2017 - April 2021) • At the Visualization Group (VisGroup), headed by Prof. Eduard Gröller • Supervisor: Ivan Viola My history at Fl MUNI • Be. 2011 - 2014, Mgr. 2014 - 2017, very different landscape • Was in the first year of Visualization course (no Vis 2) • Was in the first year of Generative design • BA with Pavel Matula (Image Processing), MA with B. Kozlikova (& I. Viola on Erasmus) • NOT the best student: • Bachelor grade average 2.17, Master grade average 2.58 • Failed a bunch of courses • But: Advisor for C Programming course (even made an exercise), Seminar lecturer for Graphics APIs (OpenGL) What it means to do a PhD • Coding prototypes of visualization software • Writing papers about it • Presenting your work • To your colleagues • To visiting scientists • At conferences 75 My PhD Projects What I'm Doing Scientific Visualization (SciVis) I Biological visualization Molecular visualization (Scientific) Visualization For Public & Exploring The Scales What is science outreach good for? • Help put things into perspective • Combine available knowledge • Hypothesis generation • Discussion starter • It's a cool looking piece of art! Interactive Visualization of Dense and Multi-Scale Data for Science Outreach David Kouřil Supervised by Ivan Viola Research Division of Computer Graphics Institute of Visual Computing & Human-Centered Technology April 1, 2021 TU Wien, Austria Rigorosum Background & Motivation ■ Bridging molecular and cellular biology results in multi-scale model MOl.rXlI.\K MODKL of i segment of cytochrome r, 11 protein ihul pi i) un important roll* in cell PMpirution, i* xhimn an it is displayed on an om-illoncope !»creen. The protein lid- III! umino i- i.l MiliiiniU; tin Moment coiimMk of unit*, ft through 1H l designated here hy llieir uhlirrviutcd names). The heme group, which aclrt a* u carrier of electron*, in known to he attached to umino ucids If and 17. In the hypolhelirul structure shown here this -Irelell »1 the molecule is assumed In lie in tin- characteristic "alpha helix" ronliguriilion. Surface containers Mesoscale^ molecular model [Johnson et al. 2014: cellPACK] [Levinthal, 1966] 15 Background & Motivation ■ GPUs enable efficient rendering of large molecular scenes [Falk et al. 2013] 16 Background & Motivation Q: How to enable people to interact with this? 19 [cellVIEW: Le Muzic et al. 2015] Background & Motivation EVERYTHING (game, 2017), David OReilly (Double Fine Productions) 21 Research Works Labels on Levels: Labeling of Multi-Scale Multi-Instance and ^ Crowded 3D Biological Environments Kouřil, L. Cmolík, B. Kozlíková, H-Y. Wu, G. Johnson, D. Goodsell, A. Olson, M. E. Gróller, I. Viola EE Transactions on Visualization and Computer Graphics, January 2019 MEMBRANE ^l^j rrymiayia|g| üyninase HyperLabels: Browsing of Dense and Hierarchical Molecular 3D Models AntiTrypsin jHiy * D. Kouřil, T. Isenberg, B. Kozlíková, M. Meyer, M. E. Gróller, I. Viola IEEE Transactions on Visualization and Computer Graphics, February 2020 many copies or me viral proteins, -i. protease cuts these viral proteins into functional sizes, and they al assemble to create new viruses. Most of the drugs currently being used to treat HI j attack these three e *B1 ro-e im i ci i u y iettectv m kTi I olecumentary: Adaptable Narrated Documentaries Using olecular Visualization \ D. Kouřil, O. Strnad, P. Mindek, S. Halladjian, T. Isenberg, M. E. Gróller, I. Viola IEEE Transactions on Visualization and Computer Graphics2pecember 2021 the virus everythin assemble: into a large cc protects the genome of Navigation Navigation in virtual environments: extensively researched Unique characteristics elicit novel approaches 3D USER INTERFACES THFDRY A M n PR AHTIP.F Navigation = Travel and Wayfinding [McCrae et al. 2009] An essential guide for anyone developing interlaces for Virtual and Augmented Heality gaining experiences. —Richard Marks, Director of Magic Lab. Sony PlayStation JOSEPH J. LaVIOLA, JR. ERNST KRUIJFF RYAN P. McMAHAN DOUG A. BOWMAN IVAN POUPYREV LaViola et al. 23 ä. Positive vs. Negative Scales positive negative Interactive Visualization for Science Outreach ■ Requires intuitive interaction methods ■ Fully interactive applications vs. pre-rendered animations ■ Explanation vs. exploration -> Exploranation [Ynnerman et al. 2018] 25 Interaction Spectrum FVLLi INTERACT! ME [Ceneda et al. 2016] "Degree of guidance" Orienting, directing, prescribing [Wohlfart & Hauser 2007] passive, playback with interactive approval, semi-interactive playback, total separation from the story 26 Navigational Degrees of Freedom Navigational Degrees of Freedom I I Visibility Annotation x> Order of exploration User-controlled System-controlled 28 ä. Modes of Navigation fXPlORAUOM ORDER CA1ER.A D ANNOTATION user-controlled AvJ&hBNTiVE- Declarative P^/roMAT^e- 29 Modes of Navigation Modes of Navigation SCIENT\FIC K/AaM6ATion> 31 J. Augmentive Navigation ■ User controls the exploration ■ However: assisted by annotation, orientation, guidance ■ Consider maps without any labels: unusable! Labels on Levels (LoL) Method David Kouřil 36 Multi-Scale Step David Kouřil 37 Multi-Scale Step to LoD for labels —► LoL capsid protein hexamer capsid HIV David Kouřil 39 Multi-Scale Step Multi-Instance Step David Kouřil 41 ■ Choose a representative instance ■ Select label anchor ■ Evaluate 4 criteria ■ Saliency criterion ■ Distance criterion ■ Border criterion ■ Temporal coherence criterion ■ Output: 2D positions of label anchors David Kouřil 42 ■ Choose a representative instance ■ Select label anchor ■ Evaluate 4 criteria ■ Saliency criterion ■ Distance criterion ■ Border criterion ■ Temporal coherence criterion ■ Output: 2D positions of label anchors David Kouřil 43 Multi-Instance Step TU W I E N Choose a representative instance Select label anchor Evaluate 4 criteria ■ Saliency criterion ■ Distance criterion ■ Border criterion ■ Temporal coherence criterion Output: 2D positions of label anchors David Kouřil 44 ■ Choose a representative instance ■ Select label anchor ■ Evaluate 4 criteria ■ Saliency criterion ■ Distance criterion ■ Border criterion ■ Temporal coherence criterion ■ Output: 2D positions of label anchors David Kouřil 45 Multi-Instance Step U E N ■ Choose a representative instance ■ Select label anchor ■ Evaluate 4 criteria ■ Saliency criterion ■ Distance criterion ■ Border criterion ■ Temporal coherence criterion ■ Output: 2D positions of label anchors David Kouřil 46 ■ Choose a representative instance ■ Select label anchor ■ Evaluate 4 criteria ■ Saliency criterion ■ Distance criterion ■ Border criterion ■ Temporal coherence criterion ■ Output: 2D positions of label anchors David Kouřil 47 Multi-Instance Step David Kouřil Labeling Step David Kouřil 49 ■ Internal vs. external labels Labeling Step David Kouřil 50 Labeling Step ■ No need for empty space ■ Weaker association beneficial ■ No leader lines overlapping David Kouřil 51 Labeling Step ■ 3D camera-aligned billboards ■ Center of the billboard = anchor point ■ Optional depth composing David Kouřil 52 Labeling Step 3D camera-aligned billboards ■ Center of the billboard = anchor point ■ Optional depth composing NOGLOBULIN David Kouřil 53 ■ Temporal coherence criterion ■ Integrated in the method ■ Biasing towards previous result ■ Label anchoring in 3D ■ Label transitions ■ Fade in, fade out ■ Travel to new position David Kouřil Implementation David Kouřil MULTI-SCALE MULTI-INSTANCE SCENE RENDERING ALGORITHM O Q) ro o Type buffer _I_ ID buffer _I_ MULTI-SCALE STEP MULTI-INSTANCE STEP Labels on Levels Method 57 Depth buffer _I I_ Color buffer i LABELING STEP Result LoL: Summary Takeaways: ■ Labeling for multi-scale, multi-instance, dense, and 3D data ■ Image-space method, real-time Initial step towards adding more non-spatial information Allows better inquiry David Kouřil Modes of Navigation VKv;kU£ATiONJ A\J6H£N/Tive K/Aa/I£AT)On> /\V/T0MATI\/£ 59 J. Modes of Navigation TU W I E N SURFACE CAPSID SCIENT\FIC receptors on the surface of cells that HIV inf bodies that block HIV infection. 60 Declarative Navigation ■ User declares the target, low-level navigation is automated ■ Good for interfacing with general audience ■ Uses metaphors from other environment: mouse clicking, links, buttons [Viola etal. 2006 TVCG] Crossover tube Piston (4) Caliper stud (2)-- Brakepad (1)-- Piston (5) Caliper stud (4) — Knuckle bracket---■ Hub----* R. caliper—" * Caliper stud (1) Dust cover Caliper stud (3) [Li etal. 2007 TOG] 61 JrTti^ii^ ^^^^ ZZZl'^ZTZ'Z.—.. . 4^ ~—— \fi_✓ upofl lha CMinaotKitiun.Tt(ii»( WM< » ^^^BIB' J ✓ [Vazquez et al. 2008 Int J CARS] ä. HOME > Plasma > 65 66 Hierarchical 3D Model Browsin Hierarchical 3D Model Browsin Hierarchical 3D Model Browsin HOME > BREADCRUMBS PANE hyperlab Hierarchical 3D Model Browsin TU WIEN HOME > brem hyperlabi Pias» HOME > Plasma > structure of interest opening Heparin AnliTrypsiii Tran^fei Aspartate Aminotransferase Fac,or " ■* Immunoglobulin A Hemoqlobin Albumin Low-Density; * Alpha-amylase Fibrinogen Low-Density Lipoprotein HIV Ceruloplasrnluymidylate Synthase fajtljsbí* Insylin *ft Immunoglobulin M 71 HyperLabels: Top-Down Navigation AntiTrypsin * * IHiy * Jmmunoglobul Immunoglobulin A # ™; Insulin Aspartate Aminotransferase Transferrin Breadcrumbs Panel: Bottom-Up Navigation » Virtual HIV HOME > Plasma > HIV > Capsid > Nucleocapsid Protein > amazon -JTTrvi Prime Electronics ~ Departments Browsing Histor 1 Coin's Amazon com Today's Deals Gift Cards & Registry Sel Help ! Computers Laptops ■ Tablets • Desktops - Monitors - Computer Accessones - PC Components - PC Gaming - Chance to win dail^ Electronics > Computers & Accessories > Computer Accessories & Pe Keyboards, Mice & Accessories > Keyboard & Mouse Combos Logitech 73 Structure of Interest Opening TU W I E N Click on label indicates interest -> select a structure of interest Triggers three procedures Sparsification of the dense model Camera anchoring to the selected substructure Re-annotation using HyperLabels describing the newly shown components All three procedures must be performed together to facilitate hierarchical 3D model browsing 74 Sparsification Activated HyperLabel ID as input, corresponds to subpart from the hierarchy Key to sparsification Leverage hierarchical organization Leverage multi-instance character Albyiin hymidylai'* Synthase Al?ma - AMYLASE 75 Sparsification Two steps: Representative instance selection: random, first occurrence, closest Visibility adjustment: all-but-one, context-cut ■ 1 76 Sparsification: Visual Style Anchoring Zooming closer to the subcomponent Embedding the camera to the local coordinate space Local navigation operations (zooming, orbiting) then relative to the subcomponents Computing bounding sphere new Re-annotation TU WIEN Deploy new HyperLabels in the scenes Update breadcrumbs pane 79 Results Limitations ■ Jumping between different parts of the hierarchy not supported ■ Conscious design decision ■ Answering the question: "How do the parts of this virus fit together?" 81 :io Modes of Navigation TU W I E N SURFACE CAPSID SCIENT\FIC K/Aa/isatiom receptors on the surface of cells that HIV inf bodies that block HIV infection. 83 Modes of Navigation SURFACE \ CAPSID few . T INTER 84 Automative Navigation User -> viewer Guided tours Scenario: public installation, no direct user-visualization interaction [Galyean 1995] [Hong etal. 1997] 85 [Andüjar etal. 2004] ä. This model includes an HIV-1 virion, capturing the virus as it is ready to infect a cells in the immune system. The virus is surrounded by blood plasma. The model is built with CellPACK, integrating many sources of information from structural biology and microscopy. The virus is surrounded by membrane. Several envelope proteins are embedded in the membrane. GP120, one of them, recognizes receptors on the surface of cells that HIV infects, and is the major target for antibodies that block HIV infection. Researchers are studying GP 120 in the search for new approaches to creating an ana-HIV vacone. f-e.e's ira enrynes ,ve pari aced rr.r.e tie vrus. Reverse transcriptase wB create a DNA copy of the genome when the virus rifects a cell. Inteqrase will splice this DNA copy into the eel's own genome, which then forces the cell to make many copies of the viral proteins. HIV protease cuts these viral proteins into functional sizes, and they all assemble to create new viruses. Most of the drugs currently being used to treat Hi; attack these three enzymes, and raraydhers c ortn ■ | even more effective drugs. Two structural proteins help Matrix protein is found on tt membrane. It plays a major the virus from the cell surfa everything is packaged msic assembles into a hexamer, which further assembles mto a large cone-shaped capsid. The capsid protects the genome of HIV, which is composed of two identical strands of P.'IA. These R'JA strands are about nine thousand nucleotides long, and encode all of the viral proteins. Nudeocapsid proteir binds to the P.'iA and helps to package it into the small space of the capsid. Together, the R'JA, the caps d. the enzymes, and the envelope work in the deadly life cyde of the virus. Models like this help researchers understand the structure and function of the virus in the search for new ways to fight infection by the virus. Engaging Explanation: Books THE MAO OF LIFE d, 1 Engaging Explanation: Molecular Animations ■ Real-time visualization instead rendered animation ■ Automated exploration instead of authored transitions ■ Synthetic commentary instead of pre-recorded voiceover 90 Molecumentary Framework: Overview STOilY <9fcApH FofcAGIMG IDs cxs NARRATIVE SYNTHESIS M ob EL STRAJCTUfcAL que m e s Rehote TEXTS STOW CUAPH Q a^Kllts P.eL*T«oiJ SHIPS O----o KEYWORD PROCESSING Molecumentary Framework: Overview stoHY £fcAi>h Fogging |D* cxs N A HRATI\/£ STOß* INPUT MOLECULAR* MobEL I structural Remote TF^TS STOW QkAPH augments KEYWORD PROCESSING FONCTIOMAL P.el*vmoiJ SHIPS o----o NAk(^ATIV£ SYNTHESIS HATCHIU 4 tb.avee.sal OF STOR.V tuei£ CD ^ Tine ^—-^W COM MEN TARY Scene P^AYfcAcK >s__^ CUT-AWA.VS CAMERA 92 Molecumentary Framework: Overview STonv^fcApH Fogging N A R.RAT I V E MOLECULAR* Mob EL i STRUCTURAL IDs cvi que fries _0 REMOTE TE /TS STOW QkAPH o----o KEYWOR.& PROCESSING NARRATIVE SYNTHESIS o-o-o o^dcr. of ST08.Y fei-£l£NTS cd ^ m COM MEN TARY ERA 93 Molecumentary Framework: Overview stoHY £fcAi>h Fogging |D* cxs N A HRATI\/£ STOß* INPUT MOLECULAR* MobEL I structural SliELEToW Remote TP * TS STOW QkAPH augments KEYWORD PROCESSING FONCTlOUM-p.EL*T»OlJ SHIPS o----o NAkfcATiVE SYNTHESIS HATCHIU 4 N/AMLAToHY tb.avee.sal of stor.v tuei£ cd ^ Ting ^—-^W COM MEN TARY Scene? P.EAL- P^AYfcAcK >s__^ CUT-AWAV5 CAMERA 94 Story Graph Foraging ■ Build basic hierarchical skeleton ■ Each object type is a node 95 Story Graph Foraging Texts describing biological components Emphasis on reusing written texts Local and remote 39 40 41 42 43 44 45 46 47 48 a 3CI»'tal-t-t|38|--iH * □ g ß u »| a n [E H s 3_ "name":"Fibrinogen", "descr":"Fibrinogen (factor I) is a glycoprotein in vertebrates that helps in the formation of blood clots." >, "root.plasma.Heparin":{ "name":"Heparin", "descr":"Heparin acts as an anticoagulant, preventing the formation of clots and extension of existing clots within the blood. While heparin does not break down clots that have already formed, it allows the body's natural clot lysis mechanisms to work normally to break down clots that have formed." } , "root.plasma.dLDL":{ "name "descr one of the five major groups of lipoproteins. Lipoproteins are complex particles composed of multiple proteins which transport all fat molecules (lipids) around "•"Low-Density Lipoprotein", "Low-density lipoprotein is Messenger RNA From Wikipedia, Ihe free encyclopedia Not to be confused with Mitochondrial DNA (mtDNA). Messenger RNA (mRNA) is a single-stranded RNA molecule that corresponds to the genetic sequence process of producing a protein. mRNA is created during the process of transcription, where the enzyme F transcript mRNA {also known as pre-mRNA). This pre-mRNA usually still contains introns, regions that wi sequence. These are removed in the process of RNA splicing, leaving only exons, regions that will encod mature mRNA. Mature mRNA is then read by the ribosome, and, utilising amino acids carried by transfer protein. This process is known as translation. All of these processes form part of the central dogma of m< genetic information in a biological system. Like in DNA, mRNA genetic information is in the sequence of nucleotides, which are arranged into codon codon codes for a specific amino acid, except the stop codons, which terminate protein synthesis. This pi acids requires two other types of RNA: transfer RNA, which recognises the codon and provides the corre (rRNA), the central component of the ribosome's protein-manufacturing machinery. The existence of mRNA was first suggested by Jacques Monod and Francois Jacob and was subsequen Matthew Meselson at the California Institute of Technology in 1961 Contents [hide] 1 Synthesis, processing and function 1.1 Transcription 1.2 Eukaryotic pre-mRNA processing 1.2.1 Splicing 1.2.2 5'cap addition 1.2.3 Editing 1.2.4 Polyadenylation 1.3 Transport 1.4 Translation 2 Structure 2.1 Coding regions 2.2 Untranslated regions i a ® 96 ä. Story Graph Foraging Establish functional relationships Alanirte arid aspartate, metab Wu etal. 2019] "Capsid protein forms a cone-shaped coat around the viral RNA, delivering it into the cell during infection." 97 Molecumentary Framework: Overview stoHY £fcAi>h Fogging |D* cxs N A HRATI\/£ INPUT MOLECULAR* MobEL I structural Remote TP * TS STOW QkAPH augments KEYWORD PROCESSING FONCTIOMAL p.EL*T»OlJ SHIPS o----o NAkfcATiVE SYNTHESIS HATCHIU 4 N/AC£AToHY TP-AVEE.SAL oiLoeR of cd ^ Tine ^—-^W COM MEN TARY Scene? P.EAL- CAMERA 98 Molecumentary Framework: Overview STonv^fcApH Fogging N A R.RAT I V E MOLECULAR* Mob EL i structural IDs cxi que fries _0 REMOTE tf/t5 STORY GRAPH aotxmeÜb^ p^Nc Tl OM A L au^c^ts RELATIONSHIPS o----o KEYWORD PROCESSING NARRATIVE SYNTHESIS o-o-o o«.DeR OF CD ^ m SCENES COM MEN TARY I $>cene ^ R EA L" T11E ^-- ERA 99 Narrative Synthesis ■ Timeline and Scenes ■ Focus scene ■ Overview scene ■ Transition scene Focus scene Transition scene 100 Scenes: Visuals Camera animation Procedural templates Anchored orbiting, direct flying, curved transition Visibility control Travelling cutting plane Exempting focused objects -mm m { m * EXEMPT wCJ™ 101 J. Scenes: Verbal Commentary TU WIEN Using Text-To-Speech library Three types ■ Structural commentary ■ Descriptive commentary ■ Navigational commentary Templating approach: ■ "$name consists of $ children" ■ "Let's look at $next" Variables based on hierarchy and tour order: H$name, $siblings, $children, $parent l$previous, $next Plasma consists for exa mple of Ceruloplasmi n, Factor H, Fibrinogen, An d others,. ™ tiTry ps in. Alpha -amy la se Structural commentary Aspartate aminotransferase is a pyridoxal phosphate (PLP)-dependent transaminase enzyme. AST catalyzes the reversible transfer of an a-ammo group between aspartate and glutamate and, as ^^^V^^* such, is an important enzyme in amino acid metabolism. Descriptive commentary Navigational commentary 102 This model ndudes an HIV-1 vrion, captunng the vrus as it is ready to infect a eels in the rrmune system. The vrus is surrounded by blood plasma. The model is bult with CelPACK, integrating many sources of information from structural biology and microscopy. The virus is surrounded by membrane. Several envelope protens are embedded in the membrane. GP120, one of them, recognizes receptors on the surface of eels that HIV infects, and is the major target for antibodies that block HIV infection. Researchers a-e studying GB120 in the search for nam approaches to creating an and-HIV vacone. Several viral enzymes are packaged inside the vrus. Reverse transcriptase wi create a DNA copy of the genome when the vrus infects a cel. Inteorase wi spice this DNA copy into the eel's own genome, which then forces the cell to make many copies of the viral protens. HIV protease cuts these viral proteins into functional sizes, and they all assemble to create new wuses. Most of the drugs currently being used to treat HI. attack these three enzymes, and researchers are - -—.;:all. r.tudj ig these enzymes to dhrx er even more effective drugs. Two structura' proteins help to organize the virus. Matrix protein s found on the nner surface of the membrane. It plays a major role during bucking of the virus from the cell surface, by making sure that everything is packaged inside. Capsid proten assembles into a hexamer, whKh further assembles nto a large cone-shaped capsd. The caps:: protects the genome of -'.. .,.":ch is composed of two identical strands of RNA. These RNA strands are about nine thousand nucleotides long, and encode al of the vral protens. fludeocapr-; protein binds to the RNA and helps to package it nto the small space of the capsid. Together, the :.'tA, the caps-d, the enzymes, and the envelope work in the deadly Ife cycle of the •JS.M0Ce: ■-- - 3 lleb JMS ■l-'-y.i-i the structure and function of the virus in the search for new ways to fight nfection by the virus. Molecumentary: Summary Framework for generating scalable documentaries from molecular models ■ Utilizes the audio track for verbal descriptions ■ Leveraging existing learning materials written by experts Interior consists for e 4kj N pie of Nucleoprotein - N termiffa£lTomain, ucledprotein - C terminal domain, Guanine a Modes of Navigation SURFACE \ CAPSID few . T INTER 106 Modes of Navigation receptors on the surface of cells that HIV ii t for antibodies that block HIV infection. SCIENTIFIC 107 J. Interactive Powers of Ten Powers of Ten (1977), Charles and Ray Eames 1 1 1 Interactive 3D Powers of Ten [Mindeketal. 2018] [Kouriletal. 2019] [Kouriletal. 2020] 108 Current Work as Post-doc: Chromatin Visualization A Brief Guide to Genomics DNA (Deoxyribonucleic Acid) C A ~ Histones Nucleosomes Gene Chromosome DNA - 1.8 m long Nucleus diameter is 10 microns (0.00001 m) 7" Cells National Human Genome Research Institute EPIGENETIC MECHANISMS are affected by these factors and processes • Development (in utero, childhood) • Environmental chemicals • Drugs/Pharmaceuticals • Aging Diet M CHROMOSOME HEALTH ENDPOINTS • Cancer • Autoimmune disease • Mental disorders • Diabetes EPIGENETIC FACTOR DNA methylation Methyl group (an epigenetic factor found in some dietary sources) can tag DNA and activate or repress genes. GENE HISTONE TAIL Histones are proteins around which DNA can wind for compaction and gene regulation. HISTONE DNA inaccessible, gene inactive DNA accessible, gene active Histone modification The binding of epigenetic factors to histone "tails" alters the extent to which DNA is wrapped around histones and the availability of genes in the DNA to be activated. Scale Light microscopy 1 000 000 - 200 nm OME-TIFF (10 GB) Super-resolution 300-20 nm OME-TIFF (0.1 GB) Electron microscopy 750 000 - 0.05 nm mmCIFOO GB) Quantitative (3C) 1000-100 nm Text(0.3 GB) Simulation (MD) 100 000-0.1 nm VTF^NG(10 GB) Representation of 3D models Hi-C All-to-all chromosome conformation capture 3D Structure • 2D contact matrices imply 3D structure • Take 2D bins as beads and use the matrix contact values as constraints -► simulate • Why it's important? • Distant regions in the sequence can be in close proximity in the spatial organization -► influences the gene expression • Genes deeper in the nucleus "hairball" might be less active • Genes deeper in the nucleus might be older in the evolution 3D genome Genome3D (multiscale) 3DGB (first person camera) HiC-3DViewer (multichromosome Chrom3D-VR TADkit (TAD visualization) GMOL (multiscale) Our idea(s) • Focus on analyzing specifically the 3D spatial structure • 3D first, not just a useless widget for decoration • Interaction with the 3D structure • Direct selection in the 3D view • Protein visualization methods applied on chromatin fiber model • SAS(A) • Radius of gyration • Shape similarity metrics So far we have this • Selecting multiple bins in 3D: important? • Vis related research question: what selections makes sense in 2D vs. 3d and vice versa • Depth in chromosome • Depth in whole genome • SASA (solvent accessible surface area) Perform section cuts and possibility to output radial distribution from the center of the system 4 r&w4 Multiscale Unfolding IEEE Vis 2021 paper Ad: Work With Us Ad: Work With Us Running research project (+ potential for future proposals and funding) Web platform (React, TypeScript, WebGPU) Potential for Bachelor and Master theses projects If interested, contact: dvdkouril@mail.muni.cz David Kouřil Bára Kozlíková Matúš Talčík Jan Byska Filip Opálený Katka Furmanová Tereza Clarence prev. PhD @ Crick Institute, London, UK Now post-doc @ ISMMS, New York, USA Thank You Bonus