src/onto/metabolomic.lisp

   1
   2 ;;; GC/MS : gas chromatograph / mass-soec
   3 ;;; data:
   4 ;;; time / ion  (2d graph)
   5
   6 (talk :name "philip dixon"
   7 ;; distance measures
   8
   9 ;; spectral fingerprints
  10 ;; peak fingerprints
  11 ;; concentrations
  12 ;; concentrations w/ ID'd compounds
  13
  14 (def data (2d :x time :y signal) )
  15
  16
  17 ( Distances:
  18   PCA of euclidean distance matrix
  19   (70+ diff distance measures)
  20
  21   (traditional-chocie: euclidane distance)
  22   (decision (or  tradition  arbitrary))  )
  23
  24 (distance measures
  25   (usually location-invariant)
  26   (this might not be true! (2->3 not same as 200->201))
  27
  28   (weighted distance measures,
  29      (incorporation mean-variance relationship)
  30      )
  31   (defun canberra-dsitance (x y) (sum (/ (abs (- y z))
  32                                          (power (- y z) 2))))
  33
  34   (:property post-weight, distance unchanges by metabolite-spec scaling)
  35
  36   (choose distance-measure
  37
  38       (important? choice)
  39       (good-chars (consistent across choice of metabolite)
  40                   ())
  41
  42       weighting seems to be more impactful than selectino of distance
  43
  44       (robustness across metabolite within distance measure)
  45           ))
  46
  47 (extensioon below-limit-of-detection, usually fixed imputation (more a LOQ than LOD case)
  48             (manage by increasing variation of measurement, in essence down-weighting))
  49
  50 (summary var-mean relationship helps selection of distance measure
  51  plus: incorporates lower precision of below-LOQ values)
  52
  53 )
  54
  55 (talk :presenter "Heike Hofmann"
  56       :title "itneractive grahics  with metabolomic data preprocessing"
  57
  58
  59       (data preproc goals
  60             (extract peaks in GC-MS)
  61             (quant amount in the peaks)
  62             (grou in metabolite profesil)
  63             (ID metaboites))
  64
  65       (realized
  66         (time vs (/ m z) ))
  67
  68       (standard sensible selection using assumption of baseline overall noise)
  69
  70       (sumamrize AUC for peak, sum for a metabolite   )
  71       (normalize across metabolutes)
  72
  73       (metabolite lookup:
  74                   NIST library ($4k)
  75                   MeT-RO coming
  76                   GOLM (not updated since 2004))
  77       (example 4 bio reps arabidopsis WT and genotype
  78
  79                (compare (old preproc Msfacts RT matching, normalized to ISTD)
  80                         (new preproc MetIDEA, log2 transform mean centered)
  81                         (chromatoplots "our method"))))