First page Back Continue Last page Overview Graphics

Conclusion

  • Based on the similarity of rankings of 21 coefficients, 13 groups can be clustered together by Kendall’s Tau.
  • Data fusion of similarity coefficients generally gives improved performance over individual coefficient.
  • Molecular size has a linear relationship with the bit density.
  • The size distribution of Simpson Coefficient behaves like that of Russell/Rao when query is small, it behaves like that of Forbes when query is large.
  • The size distributions of Cosine, Fossum, Pearson and Stiles are very similar at any query size.
  • Bit density plays an important role on chemical similarity searching.

    Notes:

    In conclusion, based on the similarity of rankings of 21 coefficients, 13 groups of coefficients can be clustered together by Kendall’s Tau. And data fusion of similarity coefficients generally gives improved performance over individual coefficient. Also, molecular sizes has a linear relationship with the bit density. The size distributions of retrieved actives of Simpson behaves like that of Russell/Rao when query is small, it behaves like that of Forbes when query is large. The size distributions of retrieved actives of Tanimoto, Baroni-Urban/Buser, Cosine and Fossum are very similar at any query size. Finally, from the mathematical proof, we see that bit density plays an important role on chemical similarity searching when overlap is considered.