368

27

Drug Discovery

such as the responses to hormones, allergens, growth signals, and so on—things

that go wrong in disease. Knowledge of the network of interactions (Sect. 23.4) is

needed to understand which proteins are the best drug targets. One hopes to develop

a physical map of the cell that will allow interpretation of masses of data through

mining techniques and will help train predictive methods for calculating pathways

and how they mesh together. Then, by homing in on the atomic details of active sites,

the best candidate drug targets—probably a very small proportion of biologically

valid targets—can be identified and subjected to closer scrutiny.

27.3 Enhancing Control of Specificity

Sophisticated drugs act by disrupting protein–protein interfaces, competing with

native binding partners. In order to do this successfully, one needs profound under-

standing of the surface of the binding site. ⁵ This knowledge can be worked into a

deep learning platform (cf. Sect. 24.3) to discover targetable epitopes. ⁶

Most therapeutically promising drugs fail because they interact with epitopes

that are structurally similar to the target epitope. This unwanted binding to often

unrelated proteins commonly produces unpredictable adverse side effects. The more

complex and coöperative the binding, the more specific it can be. Hence, mastery of

the epistructural interface, as we may call it, would lead to both very specific and

very high affinity of the drug to a unique target.

27.4 Drug–Drug Interactions

The simultaneous administration of multiple drugs can lead to therapeutic enhance-

ments. Conversely, it can also lead to adverse effects. An important motivation for

investigating therapeutically beneficial drug combinations is the possibility to, at a

stroke, vastly increase the number of diseases that can be tackled with the 4000 or so

drugs currently available. It is clearly impracticable to experimentally or clinically

test all possible drug combinations, hence any computational predictions are very

useful. ⁷ Modeling can also help to find additional uses of existing drugs. ⁸

5 Sect. 15.5.2; Fernández (2010, 2015, 2016).

6 Fernández (2019).

7 Zhang et al. (2023).

8 Abbas et al. (2021).