Bio Covalent Design

Version Compatibility

Reference examples tested with: RDKit 2024.09+, OpenEye / AutoDock Vina 1.2+ (for covalent extensions), GOLD (commercial), DOCKovalent (web service), HCovDock 1.0+.

Before using code patterns, verify installed versions match. If versions differ:

• Python: pip show rdkit then help(rdkit.Chem) to check signatures
• CLI: check version output of each docking tool

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Covalent Inhibitor Design

Design molecules that form covalent bonds with target protein residues. Clinically validated targeted covalent inhibitors include KRAS G12C inhibitors (sotorasib, adagrasib), BTK inhibitors (ibrutinib), and EGFR inhibitors (osimertinib). Covalent design requires balancing intrinsic reactivity (must form bond) vs selectivity (only the intended residue), reversibility (irreversible vs reversible covalent), and drug-likeness (warheads can hurt PK).

For warhead substructure filtering (in non-covalent contexts), see chemoinformatics/substructure-search. For non-covalent docking, see chemoinformatics/virtual-screening. For pose validation, see chemoinformatics/pose-validation.

Reactive Residue Taxonomy

Cysteine is frequently targeted because its thiol/thiolate can be nucleophilic and its local environment can support selective proximity-driven reaction. GSH and off-target cysteines are competing thiols, not intrinsically distinguishable from the target by the warhead alone; the complete ligand's recognition, exposure, and intrinsic reactivity determine selectivity.

Warhead Chemistry

Practical hierarchy: Acrylamides are common attenuated electrophiles in cysteine-directed TCIs, including KRAS G12C, EGFR, and BTK programs. Haloacetamides are generally more intrinsically reactive, but actual selectivity must be measured for the complete molecule and target context.

Decision Tree by Scenario

Kinetics: kinact / Ki

Covalent inhibition kinetics:

• Ki: reversible binding affinity (initial, like non-covalent IC50)
• kinact: rate of covalent bond formation (sec^-1)
• kinact/Ki: second-order rate constant, "covalent efficiency" (M^-1 s^-1)

For irreversible two-step inhibition that follows the corresponding kinetic model, report fitted kinact, Ki, and kinact/Ki rather than only a time-dependent IC50. Two compounds with the same IC50 can have different kinetic components:

• Low Ki, low kinact: tight binding, slow covalent bond
• High Ki, high kinact: loose binding, fast covalent bond

Because kinact/Ki depends on the target, construct, assay conditions, and kinetic model, compare values within a matched assay series and alongside exposure, intrinsic reactivity, target engagement, and selectivity. Reversible-covalent systems may require different mechanistic models and residence-time or washout measurements.

Intrinsic Reactivity Assays

Before committing to a warhead, measure intrinsic reactivity (off-target risk):

# Generic GSH stability assay readout - measure half-life of warhead with 10 mM GSH
# kinact_GSH from time-course of warhead disappearance

Do not assign a universal GSH half-life from the warhead name alone. Substitution, electronics, ionization, solubility, and assay conditions can change the observed rate. Report the GSH concentration, buffer, temperature, analytical method, and fitted half-life or second-order rate constant for the complete compound.

Covalent Docking Tools

For open-source covalent docking, HCovDock (2023) is the modern alternative; DOCKovalent is the longstanding standard.

Example: KRAS G12C Inhibitor Design Workflow

Goal: Decorate a co-crystal scaffold with a cysteine-targeting warhead and rank candidates by covalent efficiency.

Approach: Load scaffold SMILES, enumerate acrylamide-bearing analogs, filter by reactivity selectivity, dock under covalent constraint, and rank by kinact/Ki surrogates.

from rdkit import Chem

# Step 1: scaffold from co-crystal (4LRW or AMG510)
scaffold_smi = 'c1ccc(C(=O)NCC)cc1'  # generic valid scaffold for code illustration
scaffold = Chem.MolFromSmiles(scaffold_smi)

# Step 2: enumerate analogs with acrylamide warhead
def add_acrylamide(scaffold, attachment_atom_idx):
    """Project hook: attach a mapped acrylamide with an audited reaction."""
    raise NotImplementedError(
        'Provide a project-specific mapped reaction and validate atom mapping, '
        'valence, regioisomer identity, and product sanitization.'
    )

# Step 3: filter for reactive group selectivity
# Step 4: dock with DOCKovalent / GOLD covalent / HCovDock
# Step 5: rank by kinact/Ki surrogate (compute reactive Michael acceptor reactivity)

Reactivity Surrogates (computed without experiment)

For ranking warheads without wet-lab data:

Goal: Record alpha-carbon substitution as a structural feature for an acrylamide series.

Approach: Parse the SMILES, locate the acrylamide substructure, and count neighbors on the alpha carbon outside the matched warhead. This count is not a LUMO estimate or a stand-alone reactivity prediction; substituent electronics and the rest of the molecule must be considered, and reactivity should be measured.

def acrylamide_alpha_substitution_count(smi):
    mol = Chem.MolFromSmiles(smi)
    if mol is None:
        return None
    acryl_pat = Chem.MolFromSmarts(
        '[CX3:1](=[OX1:2])([NX3:3])[CX3:4]=[CX3:5]'
    )
    matches = mol.GetSubstructMatches(acryl_pat, uniquify=True)
    if not matches:
        return None
    alpha_query_idx = next(
        atom.GetIdx() for atom in acryl_pat.GetAtoms()
        if atom.GetAtomMapNum() == 4
    )
    alpha_c = mol.GetAtomWithIdx(matches[0][alpha_query_idx])
    n_subs = len([n for n in alpha_c.GetNeighbors() if n.GetIdx() not in matches[0]])
    return n_subs

For a reactivity model, use experimentally measured rates or a validated quantum-chemical workflow; a single frontier-orbital energy is not sufficient on its own.

Per-Tool Failure Modes