More on Client and Web Interfaces to RDBMS - Design and Use of Relational Databases in Chemistry - page 151

X15 0.267072 0.036639 7.289 4.63e-13 ***
X16 -0.892674 0.052076 -17.142 < 2e-16 ***
X17 0.404528 0.033359 12.126 < 2e-16 ***
X18 0.630511 0.105706 5.965 2.94e-09 ***
X19 0.316583 0.062433 5.071 4.37e-07 ***
X20 0.572682 0.160138 3.576 0.000358 ***
X21 -0.562239 0.069181 -8.127 8.05e-16 ***
X22 0.963970 0.057549 16.750 < 2e-16 ***
X23 -0.468030 0.085574 -5.469 5.15e-08 ***
X24 -0.216260 0.117513 -1.840 0.065886 .
X25 -0.241297 0.175739 -1.373 0.169909
X26 0.423008 0.079224 5.339 1.05e-07 ***
X27 0.554696 0.137816 4.025 5.93e-05 ***
X28 0.518312 0.087593 5.917 3.90e-09 ***
X29 1.375563 0.106891 12.869 < 2e-16 ***
X30 -0.224964 0.155949 -1.443 0.149320
X31 0.662818 0.187787 3.530 0.000427 ***
X32 -0.175770 0.196201 -0.896 0.370441
X33 0.825725 0.300047 2.752 0.005982 **
X34 -1.174430 0.304675 -3.855 0.000120 ***
X35 0.020394 0.371856 0.055 0.956269
---
Residual standard error: 0.6645 on 1817 degrees of freedom
Multiple R-Squared: 0.8104, Adjusted R-squared: 0.8068
F-statistic: 221.9 on 35 and 1817 DF, p-value: < 2.2e-16
This summary shows statistics on how well the counts in matrix m it the
experimental logp values. It also lists the coefficients to be used in the
above equation to estimate glogp. Finally, these coefficients, along with
the standard errors and SMARTS are stored in a data frame, dt . This data
frame is stored in the database as a table named simplex_coefficients. The
plot function produces a graph of the predicted versus experimental val-
ues as shown in Figure 12.2.
The simplex _ coefficients table can now be used in the follow-
ing SQL statement to compute a glogp value for the molecule represented
by the SMILES c1ccccc1C(=O)NC.
Select sum(contribution*count_matches('c1ccccc1C(=O)NC',smarts))
-0.262592 as glogp from simplex_coefficients
The SQL aggregate function sum and the multiplication in the above
statement effectively carry out the computation according to Formula 12.1
shown earlier. The constant value 0.262592 is the intercept from the lm fit
shown above. A glogp function can be defined as follows.
Create Function glogp(text)
Returns Numeric As $EOSQL$ Select
(sum(contribution*gnova.count_matches($1,smarts))
-0.262592)::numeric(5,3) From simplex_coefficients;
$EOSQL$ Language SQL;