Page 191 - Contributed Paper Session (CPS) - Volume 2
P. 191

CPS1820 Shuichi S.
            3.  Cancer Gene Diagnosis
                We  analyze  all  SMs  by  statistical  methods  because  all  SMs  are  small
            samples. However, we cannot obtain good results. Only RIP and H-SVM can
            discriminate all  SMs correctly.  We  realize  RipDSs,  and  H-SVM  discriminant
            scores (HsvmDSs) express the signal of microarrays. Thus, we make signal data
            using RipDSs and H-SVM DSs instead of genes. If we analyze these signal data
            by Ward cluster, it can separate two classes into two clusters. Next, we analyze
            signal data by PCA, and we propose the malignancy index by PCA in addition
            to RipDSs and HsvmDSs [10-11]. Moreover, we develop RatioSV to evaluate
            SM  and  BGS  that  is  vital  second  statistics  in  addition  to  MNM  for
            LSDdiscrimination. Now, we open the new frontier of cancer gene diagnosis
            [10-12].

            a.  Evaluation of Alon’s SMs
                We introduce the evaluation of 64 SMs found in 2017. Table 2 is two ranges
            of two classes, the range of DS, RatioSV (= 200 /DS) and two t-values. The
            range of 22 cases in class1 (Min and Max columns) are less than equal -1, and
            the range of 40 cases in class2 (MIN and MAX columns) are greater than equal
            1. SV separates two classes of 64 SMs. We consider “RatioSV of RIP” is the
            most important statistics for cancer gene diagnosis because it shows the ease
            of classification by two classes. This table is sorted in descending order of the
            value of “RatioSV of RIP.” The SV distance is two that is 26.27 %  of RipDS8's
            range. SV becomes a wide window and separate two classes completely. The
            last three rows are the maximum, mean and minimum of eight statistics. The
            range of DS, RatioSV and two tvalues are [7.47, 84.94], [2.35, 26.76], [4.22, 15.5]
            and [3.12, 14.76], respectively.

                  Table 2 The 64 RipDSs, range of DS, RatioSV (= 200 /DS) and t-values
              RipDS      Min     Max      MIN   MAX   DS       RatioSV   t (≠)    t ( = )
              RipDS8     -3.35   -1       1     4.12    7.47   26.76     15.50   14.76
              RipDS35   -2.58   -1        1     5.92    8.51   23.52     13.02   9.94
              RipDS11   -4.15   -1        1     5.52    9.67   20.68     12.71   11.17
              RipDS53   -3.72   -1        1     6.40    10.13   19.75    11.98   10.20
              RipDS27   -3.75   -1        1     6.45    10.20   19.62    13.19   11.58

              RipDS59   -15.25   -1       1     21.91   37.17   5.38     6.17    5.53
              RipDS14   -21.94   -1       1     17.85   39.79   5.03     7.16    7.42
              RipDS64   -3.94   -1        1     81.00   84.94   2.35     4.22    3.12

              MAX        -2.58   -1       1     81.00   84.94   26.76    15.50   14.76
              MEAN       -7.35   -1       1     11.69   19.04   12.84    9.49    8.63
              MIN        -21.94   -1      1     4.12    7.47   2.35      4.22    3.12

                                                               180 | I S I   W S C   2 0 1 9
   186   187   188   189   190   191   192   193   194   195   196