Imidacloprid Binder Design with AffiniBind: A Ligand-Conditioned Campaign

GenX (HFPO-DA) Binder Design with AffiniBind

A ligand-conditioned binder campaign against GenX (HFPO-DA) generated 170 candidates, with a top docking score of -7.84 kcal/mol and a target QED of 0.806.

The Target Ligand

Imidacloprid is a neonicotinoid insecticide that acts as an agonist at insect nicotinic acetylcholine receptors. Its chloropyridine and nitroimine pharmacophore make it a compact but chemically distinctive target for engineered protein binders. The molecule carries a single H-bond donor, three H-bond acceptors, and three rotatable bonds, with a molecular weight of 255.66 Da and a cLogP of 0.69. These properties place it in a favorable range for small-molecule recognition: zero Lipinski violations and a QED drug-likeness score of 0.4880. The synthetic accessibility score of 0.7556 on the 1-to-10 SAScore scale indicates that the ligand itself is straightforward to source or synthesize for downstream assays.

The ADMET aggregate risk score for imidacloprid is 0.5199, with a notably high toxicity risk score of 0.9535. That toxicity signal is consistent with the compound's known insecticidal mechanism and reinforces why a high-affinity protein binder could be useful as a detection reagent, a sequestration module, or a scaffold for engineering environmental biosensors. The campaign described here used the AffiniBind small-molecule pipeline, version 0.5.0, to generate and rank 112 candidate protein binders against this ligand.

Ligand Preparation

The ligand was prepared from the SMILES string O=[N+]([O-])N=C1NCCN1Cc1ccc(Cl)nc1, corresponding to CCD code IM4. RDKit conformer generation and partial charge assignment produced the starting geometry for pharmacophore fingerprinting and subsequent docking. The preparation stage computed the ligand-level properties reported above: molecular weight 255.66 Da, cLogP 0.69, one H-bond donor, three H-bond acceptors, three rotatable bonds, zero Lipinski violations, QED 0.4880, SAScore 0.7556, ADMET aggregate risk 0.5199, and ADMET toxicity risk 0.9535.

These values anchor the campaign's interpretation. The QED of 0.4880 is a target-level fact about imidacloprid itself, not a property of any generated binder. The same applies to the ADMET risk scores. When the pipeline later filters and ranks protein candidates, the ligand's own drug-likeness and toxicity profile remain the reference point for what the binder must recognize.

Pharmacophore Analysis

The pharmacophore fingerprinting stage identified nine features across six feature types for imidacloprid. The breakdown is:

• Acceptor.SingleAtomAcceptor: 1
• Aromatic.Arom6: 1
• Donor.SingleAtomDonor: 2
• Hydrophobe.ThreeWayAttach: 3
• LumpedHydrophobe.Nitro2: 1
• PosIonizable.Guanidine: 1

The three Hydrophobe.ThreeWayAttach features dominate the hydrophobic character of the ligand, while the single Arom6 feature captures the chloropyridine ring. The two SingleAtomDonor features and one SingleAtomAcceptor feature define the hydrogen-bonding surface that a complementary protein pocket must satisfy. The LumpedHydrophobe.Nitro2 feature marks the nitroimine group as a distinct hydrophobic element, and the PosIonizable.Guanidine feature flags the guanidine-like nitrogen environment in the imidazolidine ring. These nine features were used to condition the single-site design stage, where the ligand itself serves as the binding interface without patch discovery.

Binder Generation

The campaign generated 112 candidates total. Of these, 20 came from the boltzgen source tool and 92 from boltzgen+proofread. The generator prefilter scored 20 candidates and passed all 20; the remaining 92 candidates joined the table after the prefilter stage as variant or union rows. All 112 candidates carry a generator confidence score.

The per-tool medians show a modest separation. The 20 boltzgen candidates have a median generator confidence score of 0.6744, median solubility score of 0.7828, median SAP3D spatial score of -1.3430, median average packing density of 15.6147, and median flexible loop fraction of 0.3288. The 92 boltzgen+proofread candidates have a median generator confidence score of 0.6839. The proofread variants cluster tightly around their parent designs, which is expected given that they are refinements of the same backbone.

The top-ranked candidates table lists eight rows, but those eight rows represent four distinct designs. Six of the eight rows are variants of two parent designs: BG_SM_DENOVO_IM4_E2E_boltzgen_binder_006 and BG_SM_DENOVO_IM4_E2E_boltzgen_binder_002. The ranking reflects the parent's generator confidence score propagated to its variants, so the variant cluster should be read as a single design family rather than independent top hits.

Top Candidates

The two parent designs carry the full metric surface. The table below compares them across the dimensions where both have real measured values.

The two parent designs diverge in instructive ways. binder_006 has the stronger generator confidence (0.8318 versus 0.7977) and the better docking score (-5.68 kcal/mol versus -3.16 kcal/mol). Its SAP3D spatial score of -1.8207 is also more favorable than the -1.3984 of binder_002. Both designs are classified as soluble, with solubility scores of 0.7977 and 0.6859 respectively. Neither design has any buried unsatisfied polar atoms, and both have zero raw chain breaks and zero unplaced atoms.

The interaction profiles differ. binder_006 forms one salt bridge and zero H-bonds in the refolded PLIP analysis, while binder_002 forms three H-bonds and zero salt bridges. The delta SASA values are similar: 303.21 for binder_006 and 313.53 for binder_002. The GNINA CNN pose confidence favors binder_002 at 0.7178 versus 0.4250 for binder_006, but the CNN affinity prediction runs in the opposite direction, with binder_006 at 4.5200 and binder_002 at 5.5950. The generator's own affinity prediction value strongly favors binder_006 at 0.8583 versus 0.3416 for binder_002, though the binary probability is higher for binder_002 at 0.4116 versus 0.2534.

The six variant rows (__pf1_00, __pf1_01, __pf1_02 for binder_006, and __pf1_02, __pf1_03, __pf1_04 for binder_002) carry only their source tool, ligand gate status, and structure source. They are proofread refinements of the parent backbones, generated through the boltzgen+proofread path. Their ranking positions follow the parent's confidence score, so the cluster around binder_006 occupies ranks 1 through 4, and the cluster around binder_002 occupies ranks 5 through 8.

Docking and Developability

The docking stage used AutoDock Vina with 10 modes per candidate. The best docking score across the campaign is -5.87 kcal/mol, which appears in the full candidate set beyond the eight rows shown in the top candidates table. Among the two parent designs in the table, binder_006 achieves -5.68 kcal/mol and binder_002 achieves -3.16 kcal/mol. Both parents report their best mode as mode 1.

The developability metrics for the two parents are solid. binder_006 has an average packing density of 15.8400 and a flexible loop fraction of 0.3333, while binder_002 has a packing density of 15.8947 and a loop fraction of 0.2368. The protease exposed motif counts are low: 2.0 for binder_006 and 1.0 for binder_002. The SAP3D spatial scores of -1.8207 and -1.3984 both fall in a range that suggests acceptable spatial aggregation propensity for soluble designs.

The generator-level confidence metrics reinforce the ranking. binder_006 has a gen iPTM of 0.7794, gen pLDDT of 0.6382, and gen RMSD of 0.5951. binder_002 has a gen iPTM of 0.7521, gen pLDDT of 0.6811, and gen RMSD of 0.6802. The gen complex pLDDT values are 0.5217 and 0.5372 respectively, and the gen interaction PAE values are 9.2731 and 8.3280. The gen target PTM values are close: 0.8410 for binder_006 and 0.8365 for binder_002.

The ligand-level ADMET aggregate risk of 0.5199 and toxicity risk of 0.9535 are target properties, not binder properties. They matter for the intended application: a binder that recognizes imidacloprid must be compatible with assays where the ligand itself carries a high toxicity signal. The binder's own developability is captured by the solubility, packing, and SAP3D metrics reported above.

Validation Plan

The campaign's ordering stage produced order files for two planned experimental routes. These experiments are available and ready to run, but they have not been executed.

The first is a yeast surface display screen using IDT oPools oligo pools covering all 112 candidates, tiered by confidence, for gap-repair cloning into the pCTCON2 vector. This screen would provide the first experimental binding signal for the full candidate set.

The second is an IVTT TR-FRET assay using 96-well IDT gBlock constructs for 20 candidates. The construct context includes the T7 promoter, GG, SD, Kozak, ATG, and FLAG tag, using the E. coli MFC codon table. This assay would give a quantitative binding readout for the top tier of candidates.

The order files and protocol documents are downloadable from the target's experiments tab. Both experiments are planned and available for another lab to reproduce.

Honest Assessment

This campaign produced a workable set of ligand-conditioned binder candidates against imidacloprid, with a clear top design in BG_SM_DENOVO_IM4_E2E_boltzgen_binder_006. The pipeline's strength is the integration of generator confidence, docking, and developability metrics into a single ranking. The top parent design combines a generator confidence of 0.8318, a docking score of -5.68 kcal/mol, a solubility score of 0.7977, and a SAP3D spatial score of -1.8207. Those numbers are internally consistent: the design that the generator ranks highest also docks best and remains developable.

The main limitation is the spread between the two parent designs. binder_002 has a docking score of -3.16 kcal/mol, which is 2.52 kcal/mol weaker than binder_006. The GNINA CNN pose confidence actually favors binder_002 at 0.7178 versus 0.4250 for binder_006, which suggests that the docking pose for binder_006 may be less certain even though its score is better. The generator's own affinity prediction value of 0.8583 for binder_006 versus 0.3416 for binder_002 is a large gap, but the binary probability flips that ordering. A follow-up campaign would need to resolve this tension, likely by generating more proofread variants around binder_006 and by running the planned YSD screen to get experimental binding data.

The variant clusters are another point of caution. Six of the eight listed rows are proofread variants of the two parents, and those variant rows carry only source tool, ligand gate status, and structure source. The ranking therefore reflects the parent's score propagated to its variants, not independent evidence that each variant is a distinct top candidate. A follow-up campaign should focus on diversifying the parent backbones rather than generating more proofread variants of the same two designs.

The pharmacophore analysis identified nine features, with three Hydrophobe.ThreeWayAttach features dominating. The generated binders appear to engage this hydrophobic surface through substantial delta SASA values (303.21 and 313.53), but the H-bond and salt bridge profiles differ sharply between the two parents. binder_006 forms one salt bridge and zero H-bonds, while binder_002 forms three H-bonds and zero salt bridges. Neither design fully satisfies the ligand's two SingleAtomDonor and one SingleAtomAcceptor features with a balanced polar interaction network. A follow-up campaign would target designs that combine the salt bridge of binder_006 with the H-bond network of binder_002, ideally while preserving the stronger docking score and generator confidence of binder_006.

The ADMET toxicity risk of 0.9535 for imidacloprid itself is a reminder that the target ligand is a potent insecticide. Any binder intended for environmental sensing or sequestration must be robust enough to handle the ligand's inherent toxicity profile. The binder's own developability metrics are encouraging: both parents are soluble, have low protease exposed motif counts, and show no buried unsatisfied polar atoms. The flexible loop fractions of 0.3333 and 0.2368 are within a range that should be amenable to yeast surface display.

The planned validation experiments are the critical next step. The YSD screen with 112 candidates and the IVTT TR-FRET assay with 20 candidates are both ready to run, with order files and protocols available from the experiments tab. Until those experiments produce binding data, the computational ranking remains a hypothesis. The strongest hypothesis from this campaign is that BG_SM_DENOVO_IM4_E2E_boltzgen_binder_006 is the best imidacloprid binder among the generated candidates, with a docking score of -5.68 kcal/mol, a generator confidence of 0.8318, and a developability profile that supports experimental testing.