Emma in cdk/depict#33 suggest supporting multistep SMILES as
{A}>{B}>{C}>{D}>{E} is short for {A}>{B}>{C} and {C}>{D}>{E}. Importantly this single string allows us to depict them in a single image in CDK depict.

We already have a reaction set model in CDK so it's simply a matter of a few minor modifications to the SMILES parsing.
The depiction was a little trickier and required cleanup to how we currently handle reactions. I'm still not convinced this is that clean, but there are a lot of pieces to position in the reaction so it's always going to be a bit finiky, particularly as I wanted the ability to optimise layout for linear sequences.

Example:

ClC1=NC=2N(C(=C1)N(CC3=CC=CC=C3)CC4=CC=CC=C4)N=CC2C(OCC)=O>C1(=CC(=CC(=N1)C)N)N2C[C@H](CCC2)O.O1CCOCC1.CC1(C2=C(C(=CC=C2)P(C3=CC=CC=C3)C4=CC=CC=C4)OC5=C(C=CC=C15)P(C6=CC=CC=C6)C7=CC=CC=C7)C.C=1C=CC(=CC1)\C=C\C(=O)\C=C\C2=CC=CC=C2.C=1C=CC(=CC1)\C=C\C(=O)\C=C\C2=CC=CC=C2.C=1C=CC(=CC1)\C=C\C(=O)\C=C\C2=CC=CC=C2.[Pd].[Pd].[Cs]OC(=O)O[Cs]>C1(=CC(=CC(=N1)C)NC2=NC=3N(C(=C2)N(CC4=CC=CC=C4)CC5=CC=CC=C5)N=CC3C(OCC)=O)N6C[C@H](CCC6)O>CO.C1CCOC1.O.O[Li]>C1(=CC(=CC(=N1)C)NC2=NC=3N(C(=C2)N(CC4=CC=CC=C4)CC5=CC=CC=C5)N=CC3C(O)=O)N6C[C@H](CCC6)O>CN(C)C(=[N+](C)C)ON1C2=C(C=CC=N2)N=N1.F[P-](F)(F)(F)(F)F.[NH4+].[Cl-].CN(C)C=O.CCN(C(C)C)C(C)C>C1(=CC(=CC(=N1)C)NC2=NC=3N(C(=C2)N(CC4=CC=CC=C4)CC5=CC=CC=C5)N=CC3C(N)=O)N6C[C@H](CCC6)O>>C1(=CC(=CC(=N1)C)NC2=NC=3N(C(=C2)N)N=CC3C(N)=O)N4C[C@H](CCC4)O |f:4.5.6.7.8,16.17,18.19|  US20190241576A1

Result:

Not handled reaction/step labels yet but molecule titles/labels work well (although there is no way to get this in via SMILES:

TODO:

• John to check, new pieces are packaged-private
• SVG, px units single reaction needs tweaking