"This compound has a good docking score but low activity."
Akira was looking at the data with a puzzled expression.
"Let's look at the interaction fingerprint," Lina suggested.
"Fingerprint?" Sena asked.
"It's a representation of protein-ligand interaction patterns as a bit string."
Lina displayed it on screen. Rows of 0s and 1s.
"Bits set to 1 indicate that interaction exists."
"What kinds of interactions do you look at?"
"Hydrogen bonds, hydrophobic contacts, pi-pi interactions, salt bridges... all recorded."
Akira lined up fingerprints of two compounds.
"Compound A: 111010010110 Compound B: 110010010010"
"They're almost the same," Sena said.
"But the third bit is different. Compound A is 1, Compound B is 0."
Lina explained. "The third bit is a hydrogen bond with Asp228."
"The presence or absence of that hydrogen bond changes activity?"
"Yes. Asp228 is an important residue in the active site. Without interacting there, activity drops significantly."
Akira understood. "So that's why Compound B had low activity."
"You can't tell just from the score. You only learn from the fingerprint."
Sena wrote in her notebook. "The pattern of interactions is important."
"Yes. Even with the same score, different interaction patterns are possible."
Lina showed another example. A fingerprint of a known potent inhibitor.
"This is the ideal pattern. It satisfies all five important interactions."
"Should new compounds aim for this pattern?"
"Yes. It's close to the pharmacophore concept."
Akira supplemented. "Identify essential interaction points and find compounds that satisfy them."
"But," Sena questioned, "if it's the same pattern as a known inhibitor, doesn't it lack novelty?"
"Sharp," Lina was impressed. "That's why we maintain the core part while exploring additional interactions."
"Additional?"
"Interactions with unused pocket regions. This improves activity and selectivity."
Akira compared fingerprints. "We can calculate similarity between known drugs and candidate compounds with the Tanimoto coefficient."
"Tanimoto coefficient?"
"The number of bits commonly set to 1, divided by the total number of 1 bits."
Sena calculated. "0.75... quite similar."
"Yes. But not perfect match. We look at where they differ."
Lina pointed out the difference. "The candidate compound lacks contact with the hydrophobic pocket."
"So if we improve that..."
"Activity might increase."
Akira began drawing a new structure. "Let's try adding a hydrophobic substituent here."
Re-docking. New fingerprint.
"111011010110"
"The hydrophobic contact bit is set!" Sena rejoiced.
"Now it's closer to the known drug," Lina looked satisfied.
"Fingerprints are a summary of binding mode," Sena said.
"Yes. Representing complex three-dimensional interactions with a simple bit string."
"But," Akira warned, "fingerprints aren't perfect either."
"What do you mean?"
"Distance cutoff settings, interaction definitions... they depend on parameters."
Lina nodded. "So we look at both fingerprints and actual structures."
"Visualization is important."
Sena rotated the protein-ligand complex on screen. Hydrogen bonds displayed as yellow dotted lines.
"I see. This interaction is the key to activity."
"Fingerprints are a map. But you also have to look at the actual scenery," Lina said.
Akira concluded. "The truth told by fingerprints. Reading it is our job."
The pattern of interactions. There, the secrets of drugs are hidden.