Indian Journal of Chem istry Vol. 43A, July 2004, pp. 1369- 1 376 QSPR with TAU indices: Boiling points of sulfides and thiols KUllal Ro/ *& Achilltya Saha 2 1 Drug Th eo re ti cs a nd Cheminformatics Lab. Division of Medicinal and Pharm ace uti ca l Ch emistry De partment of Pharmaceutical T ec hnol ogy, Jada vpur University. Ko lk ata 700 032 , Indi a 2 De partment of Chemica l Techn ology , University of Calcutta, 92 , A PC Roa d, Ko lk ata 700 009 , India Ema il : kunalroy_in@yahoo.co m Recei ved 13 April 2004; revised 12 Ma y 2004 Th e prese nt paper allempts to correlate the boiling points (BP) of 26 thiols and 45 sulfid es with different, Topochemica ll y Arrived Unique (TAU) indi ces (T, T R , F, B, N v , Np, N" N x , Ny) i.n order to explore the utility of the TAU indices in modeling stud ies. Stat is ti ca l quality of the relations has bee n judge d by the parameters such as predicted variance (Q\ explained va ri ance (Rr/), co rrelation co e ffici e nt (R), variance ratio (F), etc. Th e study shows that the TAU indices can explain up to 99.7% and pred ict up \0 99.6 % of the variance of boi ling points o f the co mbined data se t. Moreove r. th e TAU indic es ca n decode spec ifi c co ntributions of molecular bulk (size), functionality, branching and shape parame ters to the boiling points of thiols and sulfides. In ge nera l. the boiling point increases with increase in bulk and d ec reases w ith in crease in branching. Further the thiol functionality contr ibutes more to th e boiling point than the sulfide functionality. Topological descriptors are formulated in graph theoretic approach 1.2 considering th e arrangements of atoms in th e (usuall y hydroge n- suppressed) molecular graph, neglecti ng three di mensional interatomic di stance s, but taking into account branching and cyclicity3-8. Various types of matrices have been used for the nume ri c description of chemical structures , e.g., adjacency matrix 5 , distance matrix 5 , reciprocal di stance matrix 9 , res ista nce distance matrix 1 0, distance h . II I' I . 12 pat matnx ,c Istance comp eme nt matnx ,reverse W · . 11 S d . L I ' . 15 lener matn x " zege matrix , ap actan matnx , B d . 16 D . 17 T I . I ur en matn x, etour matnx ,e tc . opo og lc a descriptors have been extensively used in mod e lin g different physicochemical properties (e.g., partition coefficient, water solubilit y, thermochemical properties, etc.), biological ac tivities, pharmacokinetic properties of dru gs including ADME-Toxparameters, protein binding data and bl ood- brain bar ri er partitioning, toxicity of organic chemicals and insecticid es for ecotox ic ological ri sk assess me nt etc 3 - 6 . Such modeling studies have enabled chemi sts to predict ph ys ical property or activity or toxicity of a new compound that may not have even been sy nthesized. Apart from representation of chemical compounds and reaction mecha ni sm in a graphical form, chemical grap hs find their main app li ca ti on in computing structural descriptors for QSAR/QSPR modeling and virtual scree nin g of combinato ri al Iibraries 4 • Wh ereas chemical graphs provide no n- numerical representation of chemical structure, numerical description of gra ph s, such as by me ans of matrices, is esse nti al fo r th e calculation of topological descriptors. The graph theoretical description of molecules does not consider in fo rm ation on conformational aspects, i. e., bond length s, bond angles and torsion angle s, but can encode important information on adjacency, branching a nd relative distance among different functionaliti es 111 a numerical form that determine a wide range of physicochemical and biological properti es. Significant development has been achieved in thi s fi e ld by introducing different parameters for heteroatoms and multiple bonds ap art from no ve l molecular matrices a nd graph operators. Topological descriptors have ga in ed considerable popularity as th ese can be derive d from molecular structures usi ng low computational resources. These have been considered as simple and straightfo rw ard means of viewing molecular structures as such descriptors are rapidly computa bl e from molecular structure. Though topolog ic al indices have long been used successfull y for modeling biological ac ti vit y an d physicochemical properties, recent years ha ve witnessed an upsurge in the interest of developing new graph theoretical descriptors, in view of new applications of such descriptors in similarity and diversity assessment, database mining and virtual assess me nt of comb inatorial library and pattern recognition 4. 1 8. t 9. Ve ry few topological-index-based QSPRs have been reported on the boiling points of thi ols and sulfides. Thu s, in th e present paper, we have ap pli ed