*2.1. Choice of a Model System*

We carefully chose the model systems used for database mining and quantum chemical calculations, with a focus on providing accurate representations of HBs with fluorine in the biological systems. Three types of HB donors were distinguished, namely hydroxyl, amine, and methyl group, and a pH of 7.4 ± 0.5 was considered to assess the protonation states of all entities. We extracted a ligand with interacting residue for performing high-level quantum chemical calculations in a reasonable time. This allowed us to determine the energy of isolated interaction with fluorine. For extracted amino acids from the main chain, the peptization reaction was reversed. Therefore, to maintain the proper structure of amino acids, missing atoms (hydroxyl to carboxylic group and hydrogen to nitrogen atom) were added and optimized with force field. Since 96% of the analyzed crystal structures were recorded with a resolution of <3 Å (Figure 2), we analyzed all the collected structures.

**Figure 2.** Statistical representation of the spectral resolution of the analyzed crystal structures deposited in the PDB database for L–R complexes with HB containing aliphatic fluorine (red bars) and aromatic fluorine (blue bars). "ND" (not determined) refers to the crystal structures obtained with the methods for which resolution was not specified (e.g., nuclear magnetic resonance).

### *2.2. General Statistics of HBs Containing Fluorine Atoms*

It should be emphasized that the thresholds of HB geometric parameters considered in the statistical investigation based on the PDB data can significantly influence the results and conclusion. In this study, we assumed that the HB distance was <4Å and the HB angle was 90◦–180◦. Because we aimed to determine all contacts with fluorine atoms, these values can significantly exceed the standard geometric parameters of HBs; the distance can be below the sum of the van der Waals radii of interacting atoms, and the angle can differ by up to 120◦. Complexes containing ligands with fluorine (from the LigandExpo repository) were extracted and analyzed. If the PDB entry contained more than one asymmetric unit (receptor oligomerization), the number of HBs with fluorine was multiplied by the number of occurrences of the same ligand. All measured HBs were used in further analysis, even if they came from the same PDB entry. The ligands were divided into two categories: molecules in which fluorine is bonded to an aliphatic carbon (Fal) and molecules containing fluorine bonded to an aromatic carbon (Far).

A total of 1787 (Fal) and 2324 (Far) unique PDB entries were found for fluorinecontaining ligands. Based on the assigned boundaries and defined geometric thresholds, 165 interactions with a hydroxyl group, 612 with an amine group, and 3875 with a methyl group were identified for aliphatic fluorine (Figure 3A); and 121 interactions with a hydroxyl group, 606 with an amine group, and 6698 with H–C were identified for aromatic fluorine (Figure 3C). The number of F··· H–O and F··· H–N HBs was found to be larger for Fal, whereas the number of F··· H–C HBs was two-fold higher for Far. For OH donors, three amino acids (SER, THR, TYR) were identified to be involved in HBs with Fal; however, for Far, it appeared that TYR participates less frequently in HBs, which may be attributed to the greater acidity of the OH group.

The more significant differences were observed for NH donors, in which the amino acids commonly involved in HBs with Fal were in the order ARG>ASN>GLN>GLY>LYS (Figure 3B) and in HBs with Far were in the order ARG>GLY>LYS>GLN>ASN (Figure 3D). Surprisingly, glycine is the second most common amino acid, forming HBs with Far because the others are polar amino acids with a free amino group in their side chain.

By contrast, no clear preferences of amino acids in the occurrence of HBs with Fal and Far were observed in the case of CH donors. However, the results highlighted that fluorine most frequently forms HBs with nonpolar amino acids (LEU, VAL, PHE, ILE, ALA), implying that it prefers hydrophobic areas of binding pockets (Figure 3B,D).

**Figure 3.** Number of hydrogen bonds found in the PDB repository to meet the boundaries for geometric parameters: (**A**) between Fal and OH, NH, and CH donors; (**B**) normalized division into interacting amino acids; (**C**) between Far and OH, NH, and CH donors; and (**D**) normalized division into interacting amino acids.

Based on this classification, we generated the density maps showing the geometric parameters of HBs (Figure 4). As only a small number of F··· H–O HBs were identified, certain conclusions could not be drawn (Figure 4). The density maps of HB geometric parameters obtained for the NH and CH donors (as well as OH) revealed that fluorine prefers geometries with a distance of >3 Å and an angle of 100◦–140◦ (Figure 4). However, it should be noted that more HBs were found for NH than OH donors, with a more linear geometry and short distances, but in many cases, those interactions are forced by the neighboring functional groups of a ligand interacting with amino acids.

In summary, fluorine-containing HBs reveal more angular geometric preferences than typical HBs (rather linear HBs O··· H–O, N··· H–N). Thus, in the next step, we explored the relationship between the geometry of F··· H–X (X = O, N, C) bond and the energy contribution to the ligand–receptor complexes.

**Figure 4.** Density maps of geometric parameters of HBs with (**A**) Fal and (**B**) Far.
