This is a fairly simple utility that takes a SMILES string as input and attempts to create a reasonable 3D visualization of the compound being described. To do so, it uses the following libraries:

• pyparsing, for parsing SMILES
• pyopengl, for OpenGL APIs (including GLU and GLUT)
• numpy, for linear algebra tasks

SMILES is a simple way of describing a compound by describing its constituents and how they are connected. Unfortunately, without implementing all of quantum mechanics and doing some very expensive computations, it is impossible to determine the 3D structure of a molecule using only a SMILES string. Therefore, this program uses VSEPR theory, a simple set of rules for determining 3D structure that works for many common classes of molecules. VSEPR theory is often used in the teaching of chemistry and works especially well with organic compounds.

One particular complication in the determination of 3D structure is the handling of single bonds. The relative orientations of the two atoms in a single bond can change freely because very little energy is required to twist these bonds. Therefore, compounds with single bonds can exist in many different "conformations" depending on the relative orientations of every two atoms participating in these bonds. The number of conformations increases very quickly as the number of single bonds increases. An extreme manifestation of this is in the protein folding problem, where expensive computations are needed to compute the geometries of very large molecules with many single bonds. The same problem, at a smaller scale, is also relevant to this program. For example, cyclic compounds can only exist in certain conformations.

As noted above, the program uses OpenGL for 3D rendering. However, the specific methods that it uses are somewhat outdated. Its use of GLU and GLUT in particular make it a bad representation of current OpenGL development practices. These older, simpler methods were chosen because the newer, more flexible methods were not required.

In order for all of the elements of the 3D rendering to be drawn correctly relative to each other, these elements have to be rotated in the appropriate ways. One problem was discovering how to rotate the 3D representation of a bond such that it would be pointing along a bond vector. Another problem involved rotating the 3D representation of an atom such that it would meet up with an already drawn atom. This problem was modeled as rotating a bond vector such that it would be drawn pointing in the opposite direction of an existing bond vector. The last major problem concerned the correct drawing of double bonds. The two atoms involved in a double bond will not rotate relative to each other like atoms in a single bond, so the two double bonded atoms should be drawn in a specific way relative to each other.

Despite all of the progress made, the program has limitations. The problem of single bonds, as described above, remains unsolved. In particular, solving this problem requires finding a way to represent the relative orientations of two atoms in a single bond, and there isn't a single obvious way to do that. The lack of support for cyclic compounds is another large hole in the feature-set. Solving this problem would probably involve adding the ability for geometries to deviate slightly from their VSEPR ideals, in addition to solving the single bond problem. Other miscellaneous limitations include incomplete SMILES support and a somewhat primitive UI.