(Web) Genetic Programming for Bioorthogonal Chemistry

Zeke Nierenberg

Agenda

Symbolic Regression Genetic Programming
Bioorthogonal Chemistry
Talk about 'em together!

Month	Temp
1	33
2	37
3	45
4	58
5	69
6	78
7	82
8	81
9	73
10	62
11	49
12	38

A random solution taking Month as an input, returning Temp

function(month){  return sum(pd(month,month),month);  }

Month	Temp (calc)	Temp (exp)	Δ
1	2	33	31
2	3	37	34
3	4	45	41
4	5	58	53
5	6	69	63
6	7	78	71
7	8	82	74
8	9	81	72
9	10	73	63
10	11	62	51
11	12	49	37
12	13	38	25

We breed the programs for 3 generations

function(m){ 
  return sum(m,sum(sum(0.85,mult(m,3.96)),4.94)); 
}

Ready? Now, chemistry...

Bertozzi, PNAS 2010 1821

Bioorthogonal Chemistry

Synthetically flexible
Don't react with biological functional groups
Run in water, at physiological pH
Run quickly enough to be useful
Can't be cytotoxic

Diels-Alder Cycloaddition (1950 Nobel Prize)

Cycloadditions are tempting for bioorthogonal chemistry:

Sometimes run in water
Allow for the synthesis of complex molecules
Can make things that break things (in interesting experimental ways)

We want a program that predicts the reaction rate of cycloadditions. It would take the structures of the reactants, solvent, and product as inputs. It would return a rate constant k.

function(...){ return pd(sum(solvant_cmr,pd(sum(sum(mult(dieneophile_h,solvant_cmr),pd(solvant_cmr,pd(sum(sum(mult(dieneophile_h,solvant_cmr),pd(diene_deg_unsat,subtract(dieneophile_s,sum(solvant_cl,dieneophile_cmr)))),subtract(subtract(mult(product_hba,pd(dieneophile_hbd,diene_hba)),pd(product_mw,product_mw)),pd(diene_pol_miller,temperature))),solvant_cmr))),subtract(sum(pd(diene_pol_miller,product_c),subtract(diene_cmr,dieneophile_tpsa)),subtract(pd(solvant_tpsa,subtract(solvant_xlogp2,dieneophile_xlogp2)),pd(subtract(pd(dieneophile_tpsa,sum(pd(subtract(solvant_pol_miller,solvant_s),solvant_mw),sum(subtract(diene_c,diene_h),solvant_cl))),sum(pd(sum(solvant_o,diene_c),subtract(pd(solvant_hba,dieneophile_hbd),pd(subtract(pd(dieneophile_hba,diene_h),sum(pd(solvant_hba,dieneophile_hbd),pd(subtract(diene_o,product_h),dieneophile_deg_unsat))),dieneophile_deg_unsat))),pd(diene_deg_unsat,subtract(pd(diene_pol_miller,product_c),subtract(diene_cmr,dieneophile_tpsa))))),dieneophile_deg_unsat)))),solvant_cmr)),subtract(product_c,diene_xlogp2)); }

Next Steps

Further validation
GP on spectral data
Using GP on other problems

Thank you

Prof. Lee Spector (HC)
Prof. Rayane Moreira (HC)
Prof. David Gorin (SC)
Omri Bernstein (HC)
4CBC

You can see this presentation later at: http://bit.ly/STtD0k