Some new topological indices based on the distance matrix and Randic connectivity (as graph invariants) are proposed. The calculation of these indices is simple and they have good discriminating ability toward alkanes. Incorporating the number of carbon atoms to one of the calculated indices gives a highly correlating topological index (Sh index) which found to correlate with selected physicochemical properties of wide range of alkanes, specially, their boiling points. Most of the investigated properties are well modeled (with $r^2$> 0.99) by the Sh index. Meanwhile, the resulting regressions were compared with the results based on the well-established Randic and newly reported Xu indices and, in most cases, better results were obtained by the Sh index. Moreover, multiple linear regression analysis of the alkane properties via calculated indices gives highly correlating models with low standard errors.