Molecular Tailoring Approach Fragmentation

MTA03 Fragmentation

In the context of MTA, the first fragmentation method presented was introduced by Babu and Gadre1. Using our terminology:

  1. Generate pseudoatoms (some recommendations were given including \(pi\) bonds, aromatic rings, and functional groups).

  2. Assign edges to the pseudoatoms. Let this graph be \(G_0\).

  3. Let \(L_i\) be the set of leaf pseudoatoms in \(G_i\). For each pseudoatom in \(L_i\), create a new pseudoatom which contains the original pseudoatom and the pseudoatom it is attached to. The graph resulting from compressing each pseudoatom in \(L_i\) is \(G_{i+1}\)

  4. If there are no leaf pseudoatoms in \(G_i\), form \(G_{i+1}\) by removing an edge (the original algorithm specifies that it is the first edge contained in the connectivity matrix).

  5. Repeat the previous two steps until only isolated pseudoatoms remain.

  6. Determine the actual number of atoms in each leaf pseudoatom. Let \(N_i\) be the number in the \(i\)-th pseudoatom.

  7. Compare \(N_i\) to a user-defined range. If \(N_i\) is in the range than the the \(i\)-th pseudoatom is a fragment and is removed from the graph.

  8. If pseudoatoms remain in the graph, the pseudoatoms are expanded once and the previous two steps are repeated with the resulting graph.

  9. Expansion terminates when either: all atoms are assigned to fragments, or the original pseudoatoms are recovered.

  10. In the event that the original pseudoatoms are recovered the resulting graph becomes \(G_0\) and the algorithm returns to step 2.

  11. After all atoms are assigned to fragments identify all edges which have been severed.

  12. For each severed edge grow the graph uniformally from the nodes of the edge until the resulting graph is in the target range.

The resulting set of fragments is then postprocessed. This includes:

  • Capping severed valencies with hydrogen bonds

  • Merging fragments which differ by only a few atoms

  • If weak bonds are included as edges, groups of four or less atoms connected to a larger fragment via a weak bond are removed from the fragment.

  • Close contacts between capping hydrogens are corrected by either removing a few more atoms, or by adding more atoms to make the region complete.

Note

Unlike the body of the fragmentation algorithm, the postprocessing steps are not described in detail and are thus open to some interpretation.

MTA06 Fragmentation

Along with the rebranding of MTA to CG-MTA, Ganesh et al.27 introduced a new fragmentation methodology with two user defined thresholds: a cut-off radius \(R\) and a maximum fragment size \(N\). Using our terminology:

  1. Establish pseudoatoms and connectivity

  2. Loop over pseudoatoms, for the \(i\)-th pseudoatom \(A_i\), a corresponding fragment \(F_i\) is defined as the set of pseudoatoms which are within \(R\) from \(A_i\).

  3. Merge fragments taking into account nearness and ensuring that the resulting fragments do not exceed \(N\). Exactly how to merge the fragments is not described.

    • Gadre et al.24 provides slightly more details about the merging procedure stating that it is a recursive process and that nearness and similarity are considered. Presuambly the merging is thus done by: find closest two fragments which can be merged to form a fragment of less than \(N\). Two fragments are zero distance apart if they intersect, so merge the two with the highest overlap (most atoms in common). Repeat until no two fragments can be merged and still be less than \(N\).

The manuscript also defines a concept termed “R-Goodness”. The “R-goodness” of \(A_i\) in \(F_i\) is the largest sphere which can be drawn around \(A_i\) such that all pseudoatoms in the sphere are in \(F_i\) and all pseuodoatoms outside the sphere are not in \(F_i\). In general \(A_i\) will appear in multiple fragments and the R-goodness of \(A_i\) is defined as the maximum R-goodness of \(A_i\) in any fragment. The R-goodness of the fragmentation method is the minimum R-goodness of any atom. By construction the R-goodness of the CG-MTA 06 fragmentation method is minimally \(R\) (after merging fragments it could be higher).

Note

This method was potentially introduced earlier 22; however, I am unable to access this reference to verify.

Note

This method can be viewed as the FEA Fragmentation Method with some processing steps. To our knowledge the two were developed independently however.