BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Probabilistic Pipeline and Unsupervised Learning of biomolecule ra
 ndom walks
DTSTART:20200224T140000
DTEND:20200224T150000
DTSTAMP:20260407T022025Z
UID:e3aa07cbeca3bc0cd843d2f2068c927d0c7ca3a06f5b8413cbef4b79
CATEGORIES:Conferences - Seminars
DESCRIPTION:Jean-Baptiste Masson (PhD)\n\nDecision and Bayesian Computatio
 n\, Computational biology and Neuroscience department\, Institut Pasteur\
 , Paris\, France\nTwenty years after its inception\, the field of Single M
 olecule (SM) biology undergoes a transition towards a data-generating scie
 nce [1-3]. At the nanometer scale\, the dynamics of individual biomolecul
 es is inherently controlled by random processes\, due to thermal noise and
  stochastic molecular interactions. By accessing the distribution of molec
 ular properties\, rather than simply their average value\, the great advan
 tage of SM measurements is thus to identify static and dynamic heterogenei
 ties\, and rare behaviours.\n\nIn recent years\, these experimental limits
  have been progressively alleviated with the advent of new\, game-changing
  methods. Thanks to photoactivatable probes (protein-based or synthetic dy
 es)\, millions of individual trajectories can now be recorded in live cell
 s in a few minutes. PALM/STORM images can be reliably acquired over many h
 ours (or even days)\, yielding up to hundreds of millions of individual lo
 calizations.\n\nAs SM experiments enter the age of « big data »\, the 
 development of a proper and unifying statistical framework becomes more ne
 cessary than ever.  « big data » approaches certainly open up new res
 earch venues for our understanding of biological processes\, as they enabl
 e the inference of molecular dynamics. Yet they also come with a price. Of
 ten\, adding more data brings both more information and more variability a
 nd noise. Specific tools are required to handle the complex structure of r
 esults associated to large datasets and to account for the sources of expe
 rimental and systemic variability.\n\nHere\, we show a global probabilisti
 c pipeline: TRamWAy [4-7] that automatically analyse single molecule exper
 iments from images to random walk analysis. TRamWAy relies on deep neural 
 network to deconvolve single molecule images\, Belief propagation coupled 
 to  graph summing to perform probabilistic assignments between images\, a
 nd both supervised and unsupervised Bayesian analysis to extract informati
 on from random walks.\nWe demonstrate the approach on two datasets: Glycin
 e receptors in synapses and GAG dynamics during the formation of the Virio
 n in HIV-1 [4]. We demonstrate two ways of applying the probabilistic pipe
 line TRamWAy. In the first we use model-based learning with automated resu
 lts extraction and statistics. In the second we show that unsupervised lea
 rning with structured inference allows full analysis without assigning a m
 odel to the biomolecules dynamics.\n \n[1] M. El-Beheiry et al\, Inferenc
 eMAP\,  Nat. Meth. 12\, 594–595 (2015)\n[2] M. El-Beheiry et al\,VISP\
 ,  Nat. Meth.\,  10\, pages 689–690 (2013)\n[3] InferenceMAP: h
 ttps://goo.gl/HiwoxC\n[4] Flodorer et al \, Sci. Rep. (in Press)\n[5] Re
 morino A et al\, Cell Rep 2017 Nov\;21(7):1922-1935\n[6] Knight S. C e
 t al\, Science vol 350\, Issue 6262\, p823-826 (2015).\n[7] TRamWAy: ht
 tps://goo.gl/McgJXR\n 
LOCATION:BSP 407 https://plan.epfl.ch/?room==BSP%20407
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR