International Conference on Systems Biology of Human Disease 2017

This is a live blog from the International Conference on Systems Biology of Human Disease 2017. I’ll update it throughout the conference so keep reading and follow the hashtag #SBHD17 on Twitter.

Day 1

Roland Eils w/ kickoff. Thanks to DKFZ, Harvard Medical School and University of Heidelberg.

Mike Snyder: Using omics and big data to manage health and disease

  • wearable devices for personal, time-resolved omics (genome, epigenome, transcriptome, proteome, cytokines, metabolome, lipidome, microbiome) profiling
  • see dynamic response, e.g. upon viral infections: Chen et al. Cell 2012
  • Mike Snyder saw alterations in his glucose hemoglobin HbA1c (i.e. risk factor for diabetes) levels upon viral infections and lifestyle change (be strong: He advised not to eat cookies…)
  • DNA methylation affected by nutrition and exercising as seen by whole-genome bisulfite sequencing: transcriptome bounces around like a yo-yo but methylome is a rather constant variable, which changes may indicate chronic diseases (e.g. inactivation of PDE4 by mutation and methylation involved in eosinophilia)
  • weight gain weight loss studies highlight inflammation signatures to be dynamically varying
  • wearable devices measure variables such as heart rates, skin temperature, activity patterns etc. for digital health: Li et al PLoS Biol 2017

Answers to questions: Chromatin modifications are more dynamic than DNA methylation patterns and proteome follows the transcriptome with a delay and less strong spikes. Minimum information for early diagnosis is hard to determine.

Luis Serrano: Integrative and quantitative analysis of disease networks and implications for personalized medicine

  • cancer studies serve as a paradigm for the way we analyse molecular networks of disease, which might not be applicable to other contexts
  • activations due to mutations might be propagated through network rewiring: Kiel et al. eLife 2016
  • expression of PFKM muscle and liver isoforms may opaque effect of single mutations on paralysis
  • weak pairwise correlation between Ras GEF and Ras GAP but strong correlation between sum of all Ras GEF motif-carrying molecules and sum of all Ras GAP motif-carrying molecules (GEF/GAP ratio seems constant and important to maintain healthy state; in cancer: more GEFs than GAPs)
  • competition for binding partners between network molecules: Kiel et al. Sci Signal 2013, Kiel et al. J Proteome Res 2014
  • “Edgetics” do not explain why some Ras mutations cause Rasopathy while others cause cancer. Instead, energy changes (“enedgetics”) are higher for cancer compared to Rasopathy mutations.: Kiel & Serrano Mol Syst Biol 2014
  • mutated enzymatic activity and protein folding influence each other, silent SNPs may change probabilities of such mutations
  • fragmented networks carry mutations with mild effects on essential genes affecting retina physiology and overall tissue homeostasis, but global view needs to integrate various perspectives: Kiel et al. Sci Rep 2017

Ido Amit: The power of ONE: Immunology in the age of single cell genomics

  • reference to Ilya Mechnikov, the discoverer of phagocytic cells
  • cross talk between immune cells and tissue niches influenced by nutrients and environmental factors in scope
  • FACS and microscopy aim at arriving at pure populations whereas single-cell technologies are tailored to cover broad diversity to identify populations by a data-driven approach
  • MARS-seq 2.0 scaling: 20 000 single-cell transcriptomes measured with high quality suitable for quantifications possible per day
  • challenges: longitudinal, spatial, cell-to-cell, clonal, perturbations analysis
  • ~50% of the genes expressed in liver lobule rely on the localisation of the cells: Halpern et al. Nature 2017
  • developmental pseudo-time as probed by gene editing allows to measure longitudinal single-cell transcriptomes: Mildner et al. Immunity 2017
  • disease-associated microglia (DAM) converting from the pool of homeostatic microglia in the brain
  • upregulated gene signatures in DAMs involve risk factors for Alzheimer’s disease and lead to deteriorating phagocytosis and reduced plaque clearance which drives pathological development of neurodegeneration: Keren-Shaul et al. Cell 2017

Short talks I

  • Dennis Vitkup (New York, NY, USA) on relative vs. absolute sensing of signalling systems. Another interesting paper in this respect: Frick et al. PNAS 2017
    • perceived signal is proportional to the logarithm of the fold change (i.e. relative measure) of EGFR background: Lyashenko et al. bioRxiv 2017
    • but I think: this phenomenon and the memory feature should be only conserved for absence of rapid receptor turnover (and indeed, this is the critical parameter beta)
  • Stefan Kallenberger (Heidelberg, Germany) on burst noise spectrum as Fourier transform of the activity of a popcorn machine as pattern of oscillating gene expression in virus-infected cells

Ben Lehner: Mutations and their interactions in individuals

  • investigating phenotypic effects of individual mutations in different (i.e. >1000) genotype backgrounds
  • every single mutation changes from detrimental to beneficial phenotype or vice versa in different genotypic backgrounds
  • interactions between a pair of mutations depends in the majority of cases on a third substitution across many different genotypes
  • alignment of yeast and human tRNA: 23 nucleotide substitutions, and a insertion, with 15/138 mutations switch between beneficial and detrimental

Lightning Talks I

Vincent Lotteau: Metabolic perturbations by viruses

  • pleiotropic functions of hexokinases create the need for cellular models to study the role of hexokinase (HK) isoformes in viral replication
  • pyruvate consumption of Huh-HK4 cells higher than Huh-HK2 cells with direct consequences for respiration, lipid synthesis and nucleotide synthesis through pentose phosphate pathway

Anna Marciniak‐Czochra: Mathematical modeling of clonal dynamics in acute leukemias

  • clonal hierarchy in acute myeloid leukemia (AML) cells from patient samples
  • frequencies of observed heterozygous mutations in blasts of AML cells allowed to access the combinatorial complexity of possible clonal hierarchies
  • study of evolution of clonal hierarchies from diagnosis to relapse allows in general to project, based on the current state, the time-point of relapse in the future
  • regulatory feedbacks between cells assumed as disease is far from equilibrium state
    • mechanism of negative feedback between blasts and proliferation of mitotic cells proposed as competition for growth factors and space in the niche
  • personalised cancer stem cell properties can be derived that correlate with prognosis

Day 2

Martin Howard: Analogue or Digital? Bursty or Poissonian? Dissecting the fundamentals of transcriptional regulation

  • transcription can be regulated in a digital ON/OFF fashion or in a continuously varying analogue fashion
  • epigenetic definition: inheritance of phenotype from one generation to another w/o changes in DNA sequence
  • example: by means of epigenetic maintenance through a polycomb mechanism, expression of FLC gene in Arabidopsis depends quantitatively on the length of the cold period before
  • epigenetic memory could be achieved in trans by transcription factors, which require a positive feedback, or in cis by local chromatin-based histone modification, which requires positive feedbacks, too, for propagation
  • quantitative epigenetic memory could be encoded in an analogue way (lower number of transcripts per individual cells) or digital way (switch off transcripts in individual cells while leaving them on in others)
  • establishment of histone modification patterns is a bistable process; either through competition between histone modifiers or regulated/regulating transcription factors: Berry et al. Cell Systems 2017
  • digital bistability leading to “fractional control”: switching ON/OFF individual cells digitally, which leads to an analogue fraction of cells being ON/OFF
  • epigenetic memory of FLC expression is kept locally in cis: Berry et al. eLife 2015
  • slow H3K27me3 dynamics allows chromatin state to buffer fluctuations in trans regulators
  • FLC transcription is not noisy but cell size-dependent

Naama Barkai: Visualizing gradient formation

  • morphogen sources create a concentration gradient, and gene expression is induced based on a threshold of morphogen concentration
  • canonical model of morphogen gradients depends on a local activator, but in the fly embryo the activation pattern is rather broad, which needs to be converted to sharp gradients for morphological diversification during embryogenesis
  • establishment of sharp morphogen gradients requires robustness mechanism: simple degradation and diffusion are not sufficient
  • flanking inhibitors are required: for dpp in Drosophila, there is the inhibitor Sog, which itself is inhibited by Tld
  • besides gradient of inhibition, a “shuttling” model can describe a sharp dpp gradient: a factor restricts diffusion locally as evidenced by light-sheet microscopy of Toll (receptor), Spz (ligand) and DI (morphogen)

Short talks II

  • Karsten Kuritz (Stuttgart, Germany) introduced a transformation from developmental pseudo-time to true time: Kuritz et al. J Theor Biol 2017
  • Dhana Friedrich (Berlin, Germany) is quantifying p53 dynamics and transcriptional bursts and their frequency upon induced DNA damage

Alexander Löwer: The guardian on the move: p53 dynamics and function in single cells

  • 10,000 double-strand breaks (DSBs) are induced per cell per day, which creates a frequent challenge by DNA damage
  • properties of p53 pulses: intervals between two pulses (first and last inter-pulse interval) and number of pulses
  • no deterministic description of p53 pulses depending on the number of DSBs per cell
  • delayed negative feedback in p53 regulation could give rise to Hopf bifurcation but introduction of a positive feedback rendered the system “excitable” with spontaneous cycles and sustainable oscillations
  • post-translational regulation of the p53 kinase ATM could represent positive feedbacks (by auto-phosphorylation and inhibition of inhibiting Mdm2) as transcriptional regulation does not suffice
  • hypothesised cell-to-cell variability in threshold for p53 activation controlled by Wip1: thus p53 pulses depend on both, number of DSBs and Wip1 mRNA production rate constant: Mönke et al. Sci Rep 2017
  • p53 and p21 onset correlates on population average but expression dynamics in single cells are highly variable
  • clustering revealed that p21 shows either immediate or delayed response, and it turned out that late responders are in the S phase of the cell cycle at the time of induced DNA damage

Lightning Talks II

Ana Pombo: Genome Architecture Mapping, a new approach to map chromatin contacts in rare cell types

  • regulatory contacts of distal enhancers required for activation of certain genes, which renders the global 3D chromatin architecture important
  • topologically associated domains come into focus, and even TAD-TAD contacts, which overall determine “chromosome territories” that were first visualised by FISH imaging in single cells
  • genome architecture mapping (GAM) as new technique: sequencing of DNA in thin slices of ultra-thin cryo-/laser-sectioned nuclei: Beagrie et al. Nature 2017
    • genomic regions w/ high frequency of being observed together in individual slices are likely to co-localize in the genome topology
  • length scales of chromatin contacts > 10 Mb

Kathryn Miller-Jensen: Dissecting macrophage subsets and functions with single-cell secretion profiling

  • single-cell antibody-barcoded micro-well array for secretion profiling can be applied to investigate functional plasticity of macrophages in response to lipo-polysaccharides (LPS): Xue et al. Sci Signal 2015
    • the key problem of protein microarrays remains: missing specificity and high cross-reactivity of secondary antibodies
  • tuning of polarization of tumour-associated macrophages (TAM) is a function of the tumour microenvironment
  • macrophages freshly isolated from murine melanoma tumours exhibit suppressed phenotypes only secreting a few cytokines, which would make them specific targets for immunotherapy

Fabian Theis: Lineage estimation from single-cell RNA-seq time series

  • goal is to estimate temporal dynamics from snapshots
  • diffusion maps represent multi-dimensional cellular manifold with random walks to nearest neighbours, distances of tensor can be transformed with transition probabilities as developmental pseudo-time: Haghverdi et al. Nat Methods 2016
  • merging single-cell RNA-seq datasets for efficient multi-branching:
  • series data sets can be handled by joint embedding, i.e. encoding time condition as an additional variable (projected time on tSNE map), or a dynamical system could describe the density of low-dimensional data representation at different time points
  • state-specific diffusion of cellular densities in pseudo-time described by partial differential equations w/ likelihood that allows for fitting dynamics

Barbara Treutlein: Reconstructing human cortex development using single‐cell transcriptomics

Day 3

Judith Zaugg: Variation in gene regulatory elements across individuals in health and disease

  • link single nucleotide polymorphisms (SNPs) subsequently to epigenetic alterations, transcription factor binding and mRNA levels
  • histone marks vary globally between cell types but also subtle changes come into focus
  • SNPs affect 10% of regulatory elements with histone quantitative trait loci (hQTL) and mRNA expression (e)QTL in 2kb proximity all over the genome
  • many SNPs with hQTL lay in regions of open chromatin and thus promoter regions for which transcription factor (TF) binding site scores can be calculated
  • concept: SNPs in promoter regions lead to recruitment of TFs and subsequently histone modifiers, which introduce hQTLs
    • evolutionary compensation?
  • analysis of distal SNPs showed that multiple SNPs affect a single hQTL and vice versa
  • look at SNPs from GWAS and check, which regulatory regions are affected by hQTLs
  • extensive epigenetic remodelling in pulmonary arterial hypertension (PAH) patients as compared to healthy controls but no differentially expressed genes
  • link of histone marks to TFs and discrimination between activators (positive correlation with mRNA levels) and repressors (negative correlation with mRNA levels) explained the overall low significance

Award Lectures

Short talks III

  • Helge Hass (Freiburg, Germany) showcased how a combination mechanistic mathematical modelling and machine learning can help to predict from RNA-seq data of tumours cell viability under drug administration as envisioned by Altman Cancer Discov 2015
  • Jan Baumbach (Odense, Denmark) exemplifies how a mass spectrometry-based analysis of metabolites from breath air in combination with machine learning can help to classify lung disease patients: Hauschild et al Metabolites 2015

Henrik Kaessmann: The evolution of mammalian gene expression across multiple dimensions

  • interest in the change of gene expression programs during developmental and evolutionary trajectory
  • human brain batch transcriptomes from different (st)ages can be clustered along the developmental trajectory
  • principle component analysis reveal that transcriptomes can be separated by tissue, species and developmental timing
  • neurological tissues develop quite slowly and robustly, reproduction organs develop rapidly
  • early genes during development are the least tolerance to mutations which confirms long standing hypothesis (Abzhanov Trends Genet 2013) that early in development tissues of different species are very similar as very essential and conserved gene expression programs are needed to establish the organism
  • new genes, which were emerging during evolution w/o orthologs in other species, do not really contribute to the development of the brain

Grégoire Altan‐Bonnet: Self‐organization of cytokine niches within tumors and lymphoid tissues

  • immunotherapy with higher fraction of CD4+ T cells helps to eradicate melanoma: Malandro et al Immunity 2016
  • aiming at multi-scale mathematical model for immunotherapy
  • take tumour, expose it to a short pulse of IFN-gamma to drive activation of T cells: Oyler-Yaniv et al Mol Cell 2017
    • sustained STAT1 phosphorlyation for several days
    • model revealed that a fast cytokine uptake and very slow release (i.e. “cell soaked with cytokine”) can explain long STAT1 phosphorylation
  • investigation of the spatial extent of cytokine-based communication between T cells: Oyler-Yaniv et al Immunity 2017
  • fraction of cells responding to IL2 secretion depends on the local density of cells (i.e. receptors)
    • high cell density leads to small extents of cells w/ strong response

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s