Deep learning for molecular screening · Protein–Ligand Binding Affinity Prediction

A model for one protein target

Screening asks which compounds to prioritise from a collection. In our study, a target-specific model learns from compounds labelled for one protein. It uses the ligand’s sequence and geometry to rank compounds by predicted activity against that target.

Protein–compound pair prediction on the left and target-specific compound screening on the right. — **Target-specific screening.** Pair-based models use protein and compound features together. Target-specific models learn to recognise candidate compounds for one protein. Figure 1, Ke and Wang (2026).

Why use two descriptions?

A SMILES sequence records chemical connectivity, while a molecular grid provides spatial context. Our question was whether learning interactions between these descriptions improves the ranking of active compounds.

DL-FSG: learning across two molecular views

Deep learning on the fusion of chemical sequences and molecular grids for ligand-based virtual screening ↗
Rongji Ke and Debby D. Wang
BIO Web of Conferences 232, 02002 (2026)

DL-FSG encodes the SMILES sequence and ligand grid in parallel. Bidirectional cross-attention lets the features from each branch inform the other. The combined features then produce a target-specific screening score.

DL-FSG architecture with parallel SMILES and molecular grid encoders, bidirectional cross-attention and a screening prediction head. — **Sequence–grid fusion.** DL-FSG combines sequence and spatial features through cross-attention. Figure 2, Ke and Wang (2026).

Screening across 102 DUD-E targets

DL-FSG distinguishes active compounds from decoys with an AUROC of 0.9910 and an AUPRC of 0.9725 on the DUD-E benchmark.

AUC and AUPR comparisons between DL-FSG and baseline methods across 102 DUD-E targets. — **Screening performance across targets.** Screening performance across 102 DUD-E targets. Figure 3, Ke and Wang (2026).

What changes when a component is removed?

Replacing cross-attention with self-attention reduced AUC to 0.9855. Using either SMILES or grids alone also reduced performance. The two ablation studies support learning from both representations and their interaction.

Attention comparison: cross-attention AUC 0.9910 and AUPR 0.9725; self-attention 0.9855 and 0.9665; no attention 0.9798 and 0.8448. — **Attention comparison.** Cross-attention improves both AUC and AUPR over self-attention and the configuration without attention. Table 4, Ke and Wang (2026).

Representation comparison: combined features AUC 0.9910 and AUPR 0.9725; SMILES alone 0.9683 and 0.8323; grids alone 0.9646 and 0.8266. — **Representation comparison.** Combining sequence and grid features improves screening performance over either representation alone. Table 5, Ke and Wang (2026).

Reading the grid branch with Grad-CAM

Grad-CAM highlights spatial regions associated with a higher screening score. It uses score gradients to weight the grid branch’s feature maps. The gradients pass through the sequence–grid fusion model while the molecule and conformer remain fixed. [2]

Ethanol conformer and the blue plane at the z coordinate of the Grad-CAM cross-section. — **The response extends across the plane.** This ethanol example produces a broad Grad-CAM response. Atom markers locate the molecule within the map, whose shading shows positive Grad-CAM intensity. Grad-CAM: Selvaraju et al. (2017), applied to DL-FSG.

Ethanol Grad-CAM cross-section on a fixed zero to 0.00001 scale, with projected heavy-atom positions and a broad response across the plane. — **The response extends across the plane.** This ethanol example produces a broad Grad-CAM response. Atom markers locate the molecule within the map, whose shading shows positive Grad-CAM intensity. Grad-CAM: Selvaraju et al. (2017), applied to DL-FSG.

Each voxel summarises a spatial neighbourhood, giving a coarse view of the model’s response. Read the map at this scale rather than as a contribution from an individual atom. A shared colour scale allows comparisons across molecules.

Screening analysis

Try the MAO-B screening model with an example molecule or your own SMILES sequence. Its score ranks molecules for this target; it is not a calibrated probability of activity. After calculating a score, choose Calculate Grad-CAM to explore the spatial response.

Preparing analysis…

References

Rongji Ke and Debby D. Wang. Deep learning on the fusion of chemical sequences and molecular grids for ligand-based virtual screening. BIO Web of Conferences 232, 02002 (2026).
Ramprasaath R. Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh and Dhruv Batra. Grad-CAM: Visual Explanations From Deep Networks via Gradient-Based Localization. Proceedings of the IEEE International Conference on Computer Vision, pp. 618–626 (2017).

Screening