DDE: A Chlorinated Aromatic Target for Ligand-Conditioned Binder Design

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

DDE is a chlorinated aromatic compound with the SMILES string ClC(Cl)=C(c1ccc(Cl)cc1)c1ccc(Cl)cc1. It carries the CCD code 6WS and appears as a co-crystal ligand under the PDB ID DDE. The molecule is built around a dichloroethylene core flanked by two para-chlorophenyl rings, giving it a compact, hydrophobic, and heavily halogenated architecture. With a molecular weight of 318.03 Da and a cLogP of 6.19, DDE sits at the lipophilic edge of drug-like chemical space. It has zero hydrogen bond donors and zero hydrogen bond acceptors, two rotatable bonds, and one Lipinski violation. Its QED drug-likeness score is 0.6193, its synthetic accessibility score is 0.8989 on the SAScore scale where 1 is easy and 10 is hard, and its ADMET aggregate risk score is 0.5430 with a toxicity risk score of 0.8370.

The ligand matters as a design target because its dense halogenation and aromatic surface create a distinctive recognition problem. A protein binder must complement a ligand that offers no conventional hydrogen bonding handles, relying instead on shape complementarity, van der Waals contacts, and halogen-adjacent packing. The campaign described here used the AffiniBind small molecule pipeline, version 0.5.0, to generate and rank 120 candidate protein binders against DDE.

Ligand Preparation

The ligand preparation stage began with the DDE SMILES string and the CCD code 6WS. RDKit conformer generation produced the three-dimensional starting geometry, and partial charges were assigned for downstream docking and interaction scoring. The prepared ligand then fed directly into pharmacophore fingerprinting and the single-site design stage, where the ligand itself served as the binding interface. No patch discovery was required because the small molecule pipeline treats the ligand as the anchor point for binder generation.

The ligand-level properties computed once for DDE frame the entire campaign. The molecular weight of 318.03 Da and cLogP of 6.19 place DDE in a hydrophobic regime where solubility and metabolic clearance are expected to be challenging. The ADMET aggregate risk score of 0.5430 and the toxicity risk score of 0.8370 reinforce that concern. The synthetic accessibility score of 0.8989 indicates that DDE itself is straightforward to synthesize, which is useful for experimental follow-up. The QED score of 0.6193 reflects the tension between the ligand's compact size and its high lipophilicity.

Pharmacophore Analysis

The pharmacophore fingerprinting stage identified 18 pharmacophore features for DDE. The feature distribution breaks down into three categories: 2 aromatic features of type Aromatic.Arom6, 14 hydrophobic features of type Hydrophobe.ThreeWayAttach, and 2 lumped hydrophobic features of type LumpedHydrophobe.RH6_6.

The dominance of hydrophobic features is consistent with the ligand's structure. The two para-chlorophenyl rings and the dichloroethylene core present a largely nonpolar surface, and the 14 three-way-attach hydrophobic features capture the branched, multi-ring topology. The two aromatic features mark the electron-rich ring systems, while the two lumped hydrophobic features summarize the larger hydrophobic patches. This pharmacophore profile guided the conditioning signal for binder generation, emphasizing shape and hydrophobic complementarity over polar interaction motifs.

Binder Generation

The binder generation stage produced 120 candidates in total. Of these, 20 candidates passed the generator prefilter, and 0 candidates were excluded by that prefilter. The remaining 100 candidates joined the table after the prefilter stage as variant or union rows, meaning they were never screened by the prefilter itself. The source-tool split is clean: 20 candidates came from boltzgen alone, and 100 candidates came from boltzgen+proofread.

All 120 candidates carry a generator confidence score. The median generator confidence score for the 20 boltzgen candidates is 0.6054, and the median for the 100 boltzgen+proofread candidates is also 0.6054. The boltzgen candidates additionally carry median solubility score of 0.7839, median SAP3D spatial score of -1.3918, median average packing density of 15.0623, and median flexible loop fraction of 0.3173.

The top candidates table lists 8 rows, but those rows represent only 4 distinct designs. Six of the listed rows are variants of 2 parent designs. The parent design BG_SM_DENOVO_6WS_E2E_boltzgen_binder_001 appears with five proofread variants, and the parent design BG_SM_DENOVO_6WS_E2E_boltzgen_binder_002 appears with one proofread variant. The ranking by generator confidence score reflects the parent's score propagated to its variants, so the variant cluster around binder_001 occupies the top five positions after the parent itself.

Top Candidates

The ranked table below lists the candidates that carry generator confidence scores and docking scores. The variant rows carry only the metrics listed on their own lines, so the table is restricted to columns where every listed row has a real value.

The top ranked candidate, BG_SM_DENOVO_6WS_E2E_boltzgen_binder_001, carries a generator confidence score of 0.7735 and a docking score of -5.86 kcal/mol. Its GNINA CNN pose score is 0.6906 and its GNINA CNN affinity value is 5.0140. The second parent design, BG_SM_DENOVO_6WS_E2E_boltzgen_binder_002, has a generator confidence score of 0.7068 and a docking score of -6.78 kcal/mol, with a GNINA CNN pose score of 0.7284 and a GNINA CNN affinity value of 5.7220.

The five proofread variants of binder_001 and the single proofread variant of binder_002 carry no generator confidence scores or docking scores on their own lines. They are ranked in the top candidates table by their parent's confidence score, but their individual metric values are limited to the source tool, ligand gate status, and structure source fields. The variant cluster around binder_001 occupies ranks 2 through 6, and the binder_002 variant occupies rank 7.

Docking and Developability

The campaign's best docking score is -9.46 kcal/mol, which appears in the full campaign ranking but not on the two parent rows shown in the top candidates table. The two parent designs that do carry docking scores show -5.86 kcal/mol for binder_001 and -6.78 kcal/mol for binder_002. Both parents were docked with 10 modes, and both had their best mode at mode 1.

The developability metrics for the two parent designs are substantial. Binder_001 has a solubility score of 0.7440 with a solubility class of soluble, a SAP3D spatial score of -1.9914, an average packing density of 14.2222, one protease exposed motif, zero buried unsatisfied polar atoms, and a flexible loop fraction of 0.1587. Binder_002 has a solubility score of 0.8613 with a solubility class of soluble, a SAP3D spatial score of -1.2489, an average packing density of 15.1667, two protease exposed motifs, zero buried unsatisfied polar atoms, and a flexible loop fraction of 0.3194.

The generator-level structural metrics for binder_001 include a gen_iptm of 0.7244, a gen_plddt of 0.7570, a gen_rmsd of 0.6357, a gen_ligand_iptm of 0.7244, a gen_min_interaction_pae of 1.6865, a gen_complex_iplddt of 0.6321, a gen_interaction_pae of 8.1579, a gen_complex_pde of 0.7454, a gen_ptm of 0.7888, and a gen_target_ptm of 0.8034. The refolded structure shows zero hydrogen bonds and zero salt bridges to the ligand, a delta SASA of 391.4005, an affinity prediction value of 0.9196, and an affinity probability binary of 0.3987.

Binder_002 shows a gen_iptm of 0.6754, a gen_plddt of 0.7280, a gen_rmsd of 0.8576, a gen_ligand_iptm of 0.6754, a gen_min_interaction_pae of 2.3546, a gen_complex_iplddt of 0.6402, a gen_interaction_pae of 8.8179, a gen_complex_pde of 1.0482, a gen_ptm of 0.7816, and a gen_target_ptm of 0.8004. Its refolded structure shows zero hydrogen bonds and zero salt bridges to the ligand, a delta SASA of 408.4924, an affinity prediction value of 1.5115, and an affinity probability binary of 0.3356.

The ADMET assessment for this campaign is a target-level property, not a per-candidate metric. DDE itself carries an ADMET aggregate risk score of 0.5430 and a toxicity risk score of 0.8370. The QED drug-likeness score of 0.6193 and the synthetic accessibility score of 0.8989 are also target-level values. These numbers describe the ligand, not the protein binders, and they frame the developability context for any downstream application.

Validation Plan

The campaign's ordering stage produced order files for two planned experiments. The first is a yeast surface display screen using IDT oPools oligo pools covering all 120 candidates, tiered by confidence, for gap-repair cloning into the pCTCON2 vector. The second is an IVTT TR-FRET assay using 96-well IDT gBlock constructs for 20 candidates, with the T7 promoter plus GG plus SD plus Kozak plus ATG plus FLAG construct context using the E. coli MFC codon table.

These experiments are planned and available for reproduction. They have not been run, and no Kd values, titration curves, or other experimental results exist. The order files and protocol documents are downloadable from the target's experiments tab.

Honest Assessment

This campaign produced a coherent set of ligand-conditioned binder candidates against DDE, but the results expose a clear tension between generator confidence and docking validation. The top ranked parent design, binder_001, carries the highest generator confidence score at 0.7735, yet its docking score of -5.86 kcal/mol is weaker than the second parent design's -6.78 kcal/mol. The campaign's best docking score of -9.46 kcal/mol appears elsewhere in the full ranking, which suggests that the generator confidence ranking and the docking ranking are not aligned. A follow-up campaign should weight docking scores more heavily in the final ranking, or apply a consensus scoring approach that combines generator confidence, docking score, and GNINA CNN pose confidence.

The pharmacophore profile of DDE is dominated by hydrophobic features, and the generated binders reflect that bias. Both parent designs show zero hydrogen bonds and zero salt bridges to the ligand in their refolded structures, relying entirely on shape complementarity and van der Waals packing. The delta SASA values of 391.40 and 408.49 square angstroms indicate substantial burial of surface area, which is encouraging for binding affinity, but the absence of polar contacts may limit specificity. A follow-up campaign could introduce polar residues near the chlorinated rings to test whether halogen bonding or weak polar interactions improve selectivity.

The developability metrics for the two parent designs are acceptable. Both are soluble, both have zero buried unsatisfied polar atoms, and both have SAP3D spatial scores in a reasonable range. Binder_001 has a lower flexible loop fraction at 0.1587 compared to binder_002's 0.3194, which may favor stability. The protease exposed motif counts are low, at one and two respectively. These are workable starting points for experimental validation.

The variant strategy is the weakest part of the current data surface. The 100 proofread variants carry no generator confidence scores, no docking scores, and no developability metrics on their own lines. They are ranked by their parent's confidence score, which means the top candidates table overrepresents a single design. The campaign generated 120 candidates, but the table's 8 rows represent only 4 distinct designs, and 6 of those rows are variants of one parent. A follow-up campaign should either score the variants independently or report the ranking at the design level rather than the variant level.

The planned validation experiments are well specified. The yeast surface display screen covers all 120 candidates, and the IVTT TR-FRET assay covers 20 candidates with a defined construct context. These experiments are ready to run and available for another lab to reproduce. The order files and protocol documents are downloadable from the experiments tab. The next step is experimental execution, and the results will determine whether the generator confidence scores or the docking scores better predict actual binding against DDE.