A ligand-conditioned binder design campaign against perfluorobutanesulfonamide (PFBSA) generated 198 candidates, with a top docking score of -6.80 kcal/mol and a target-level QED of 0.806.
Explore more from this discipline or browse the full archive.
A ligand-conditioned design campaign against perfluorobutanesulfonamide generated 195 binder candidates, with a top docking score of -7.74 kcal/mol and a target QED of 0.806.
Send us a target surface and we will scope a feasibility review, design binders, and report the same objective metrics our research is built on.
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.
A full AffiniBind campaign against perfluorononanoic acid generated 194 binder candidates, with the top design reaching a docking score of -9.50 kcal/mol and a generator confidence of 0.9549.
PFBSA, or perfluorobutanesulfonamide, is a small, heavily fluorinated sulfonamide with the SMILES string NS(=O)(=O)C(F)(F)C(F)(F)C(F)(F)C(F)(F)F. The molecule carries a single sulfonamide head group attached to a four-carbon perfluoroalkyl tail. That tail is the defining feature: four contiguous CF2 groups terminated by a CF3, creating a short, rigid, highly hydrophobic fluorocarbon chain. The sulfonamide end is the only polar handle, offering one hydrogen-bond donor and two acceptors. The ligand is compact, with a molecular weight of 299.11 Da, three rotatable bonds, and a computed cLogP of 1.70. It violates none of the Lipinski rules.
Why target PFBSA with a designed protein binder? The perfluoroalkyl chain is chemically inert and metabolically persistent, while the sulfonamide head group provides a modest but real anchoring point for molecular recognition. A protein pocket that can discriminate PFBSA from other perfluorinated sulfonamides, or from hydrocarbon sulfonamides, would need to exploit both the polar sulfonamide and the unusual fluorocarbon surface. The campaign described here used the AffiniBind small molecule pipeline to generate and rank candidate protein binders against this ligand, starting from the ligand structure alone.
The ligand preparation stage began with the PFBSA SMILES string and the CCD code KPQ. RDKit conformer generation produced the three-dimensional starting geometry, and partial charges were assigned for downstream docking and interaction scoring. The prepared ligand was then carried into pharmacophore fingerprinting and binder generation.
The target-level properties computed once for PFBSA are listed in the table below. These values describe the ligand itself, not any individual protein candidate.
| Property | Value |
|---|---|
| Molecular weight | 299.11 Da |
| cLogP | 1.70 |
| H-bond donors | 1 |
| H-bond acceptors | 2 |
| Rotatable bonds | 3 |
| Lipinski violations | 0 |
| QED drug-likeness | 0.806 |
| Synthetic accessibility (SAScore) | 0.783 |
| ADMET aggregate risk score | 0.507 |
| ADMET toxicity risk score | 0.901 |
The QED value of 0.806 places PFBSA in a favorable drug-likeness range, driven by its low molecular weight and modest lipophilicity. The synthetic accessibility score of 0.783 on the SAScore scale, where 1 is easy and 10 is hard, indicates a readily synthesizable ligand. The ADMET aggregate risk of 0.507 sits near the midpoint of the risk scale, while the toxicity risk component is elevated at 0.901, a signal that the perfluorinated chain may carry toxicological liabilities. These ligand-level properties frame the target as a small, synthetically accessible molecule with a mixed ADMET profile.
The pharmacophore fingerprinting stage identified three features on PFBSA: two single-atom acceptors and one single-atom donor. These features map directly to the sulfonamide group. The two oxygen atoms of the sulfonyl moiety contribute the acceptor features, and the sulfonamide nitrogen contributes the donor feature. The perfluoroalkyl tail contributes no pharmacophore features, consistent with its chemically inert, nonpolar character.
This three-feature pharmacophore is sparse but chemically coherent. The entire polar recognition surface of PFBSA is concentrated in the sulfonamide head group. A designed protein pocket must therefore present complementary hydrogen-bonding residues to engage the sulfonamide, while the fluorocarbon tail will be accommodated largely through shape complementarity and van der Waals contacts. The pharmacophore fingerprint, combined with the Morgan and MXFP fingerprints and GNN embeddings from molfeat, provided the conditioning signal for the binder generation stage.
The campaign generated 198 candidates using the AffiniBind pipeline version 0.5.0. The source-tool split shows 80 candidates from the boltzgen generator and 118 candidates from the boltzgen+proofread path. All 198 candidates carried a generator confidence score. The generator prefilter passed 80 candidates and excluded 0 candidates. The 118 candidates that joined the table after the prefilter stage are variant or union rows, never screened by the prefilter.
The single-site design strategy treated the ligand as the binding interface, with no patch discovery step. The protein binder generation used PXDesign and BoltzGen, conditioned on the ligand structure to produce complementary protein pockets. The proofread variants represent a second pass over selected parent designs, refining the generated sequences and structures.
The per-tool median generator confidence scores provide a coarse view of the two source populations. The 80 boltzgen candidates had a median confidence of 0.698. The 118 boltzgen+proofread candidates had a median confidence of 0.807. The proofread population shifted the confidence distribution upward by roughly 0.11, consistent with the proofreading step selecting and refining higher-confidence parent designs.
The ranked table below lists the eight candidate rows that appear in the top candidates block. These eight rows represent three distinct designs: one parent design with six variant rows, and a second parent design with two variant rows. The ranking reflects the parent's generator confidence score propagated to its variants, so the variant cluster from the top parent occupies the first six rows.
| Candidate | Source tool | Generator confidence | Docking score | CNN pose score | CNN affinity |
|---|---|---|---|---|---|
| BG_SM_DENOVO_KPQ_L80B_E2E_boltzgen_binder_033 | boltzgen | 0.916 | -6.03 kcal/mol | 0.689 | 4.094 |
The top parent design, BG_SM_DENOVO_KPQ_L80B_E2E_boltzgen_binder_033, carries a generator confidence score of 0.916, the highest in the listed set. Its docking score is -6.03 kcal/mol, with a CNN pose score of 0.689 and a CNN affinity value of 4.094. The six proofread variants of this parent share the same docking score, CNN pose score, and CNN affinity value, reflecting their common structural origin. The second parent design, BG_SM_DENOVO_KPQ_L80B_E2E_boltzgen_binder_014, appears through two proofread variants, also with docking scores of -6.03 kcal/mol and CNN pose scores of 0.689.
The parent design BG_SM_DENOVO_KPQ_L80B_E2E_boltzgen_binder_033 also carries a set of structural and developability metrics. Its solubility score is 0.464, classified as insoluble. The SAP3D spatial score is -1.282, the average packing density is 15.070, and the flexible loop fraction is 0.296. The design has one protease-exposed motif and zero buried unsatisfied polar atoms. The generator-reported interface metrics include an iPTM of 0.860, a pLDDT of 0.661, an RMSD of 0.542, and a ligand iPTM of 0.860. The interaction PAE is 7.392, and the complex pLDDT is 0.529. The PLIP analysis of the refolded structure reports zero hydrogen bonds and zero salt bridges, with a delta SASA of 275.054 Ų. The affinity prediction value is 2.043, with a binary affinity probability of 0.131.
The docking stage used AutoDock Vina with 10 modes per candidate. The best docking score across the campaign was -6.80 kcal/mol. The top listed parent design docked at -6.03 kcal/mol, a moderate score that suggests a plausible but not exceptionally strong predicted binding pose. The CNN pose score of 0.689 indicates a reasonably confident pose assignment, while the CNN affinity value of 4.094 sits in a range that the Gnina model associates with moderate predicted affinity.
The developability metrics for the top parent design paint a mixed picture. The solubility score of 0.464 with an insoluble classification is a concern for downstream experimental work, since an insoluble binder is difficult to express and assay. The SAP3D spatial score of -1.282 is negative, which in the SAP3D framework points toward spatial aggregation propensity. The average packing density of 15.070 and the flexible loop fraction of 0.296 are within typical ranges for designed proteins. The single protease-exposed motif and zero buried unsatisfied polar atoms are favorable: the design avoids internal polar strain and has limited obvious proteolytic vulnerability.
The generator-reported interface metrics for the top parent are more encouraging. The iPTM of 0.860 and the ligand iPTM of 0.860 indicate that the generator is confident in the interface geometry. The interaction PAE of 7.392 is low, suggesting well-resolved relative positioning between the protein and ligand. The complex pLDDT of 0.529 is modest, reflecting some uncertainty in the overall complex structure. The delta SASA of 275.054 Ų upon refolding indicates substantial burial of surface area at the interface, consistent with a real binding event. The absence of PLIP-detected hydrogen bonds and salt bridges in the refolded structure is notable: the designed interface appears to rely on shape complementarity and van der Waals contacts rather than polar interactions, despite the sulfonamide pharmacophore offering two acceptors and one donor.
The affinity prediction value of 2.043 and the binary affinity probability of 0.131 suggest that the generator's own affinity model is not strongly convinced this design will bind with high affinity. This is a cautionary signal, though the docking score and interface metrics provide some counterweight.
The campaign's ordering stage produced order files for two planned experiments. These experiments are available and ready to run, but they have not been executed.
The first planned experiment is a yeast surface display screen using IDT oPools oligo pools covering all 198 candidates, tiered by confidence, for gap-repair cloning into the pCTCON2 vector. This screen would provide a direct experimental readout of binding for the full candidate set.
The second planned experiment is an IVTT TR-FRET assay using 96-well IDT gBlock constructs for 80 candidates. The constructs use the T7 promoter with a GG + SD + Kozak + ATG + FLAG construct context, designed for the E. coli MFC codon table. This assay would provide a quantitative binding signal for the top tier of candidates.
The order files and protocol documents are downloadable from the target's experiments tab: /targets/small-molecule/pfbsa-perfluorobutanesulfonamide/experiments.
This campaign produced a coherent set of ligand-conditioned binder candidates against PFBSA, with a clear top parent design and a well-defined variant cluster. The pipeline ran from ligand preparation through pharmacophore fingerprinting, binder generation, docking, and ADMET assessment without losing candidates to the prefilter: all 80 boltzgen candidates passed, and the 118 proofread variants joined the table afterward. The top parent design achieved a generator confidence score of 0.916 and a docking score of -6.03 kcal/mol, while the campaign-wide best docking score reached -6.80 kcal/mol.
What worked well is the pharmacophore analysis. The three features identified on PFBSA, two acceptors and one donor, correctly capture the sulfonamide head group as the sole polar recognition element. The binder generation stage produced a ranked set with a clear top parent and a proofread variant cluster that preserves the parent's docking and CNN metrics. The interface metrics for the top parent, particularly the iPTM of 0.860 and the interaction PAE of 7.392, indicate that the generator is confident in the ligand placement.
What the pipeline struggled with is the developability profile of the top parent. The solubility score of 0.464 with an insoluble classification is a real liability for experimental validation. The SAP3D spatial score of -1.282 reinforces that concern. The affinity prediction probability of 0.131 is low, and the refolded structure shows no PLIP-detected hydrogen bonds or salt bridges despite the sulfonamide pharmacophore offering clear polar handles. The design appears to bury substantial surface area, with a delta SASA of 275.054 Ų, but the absence of polar contacts suggests the interface may rely too heavily on shape complementarity against the fluorocarbon tail.
A follow-up campaign would target the solubility and polar engagement problems directly. The sulfonamide head group should be anchored with explicit hydrogen-bonding residues, and the fluorocarbon tail should be accommodated in a more enclosed hydrophobic pocket to improve both affinity and solubility. The proofread variants, which share the parent's docking score and CNN metrics, offer a natural starting point for sequence-level optimization aimed at improving solubility and introducing polar contacts. The planned YSD screen and IVTT TR-FRET assay, once run, will provide the experimental binding data needed to distinguish the top candidates and guide the next design cycle.