Drug chemistry is complicated: simple patterns are not the norm
For a drug to be effective it must meet many criteria, some simple and some very complex.
Bioavailability, for example, depends on the route of delivery and even this is sometimes complicated. Orally delivered drugs (eg analgesic pills) have to survive the stomach and also have to be absorbed into the body from some part of the digestive system. This may not be easy to achieve. Sometimes even small changes in a molecule will vastly alter its absorption, stability in the digestive tract or active concentration in the body. Sometimes small changes may make the molecule far easier for the body to metabolise, greatly lowering the bioavailability. And those things are the simple side of the problem.
Many drugs work by inhibiting specific enzymes. Here the problem gets even more complicated. Good inhibitors need to match and bind to an active site in a specific enzyme. Sometimes similar molecules will have some activity for a specific target and similar variants will have similar activity. For example, there are structural similarities between the commonest NSAIDs, paracetamol (acetaminophen), aspirin, ibuprofen and naproxen (they all inhibit the cyclooxygenase enzymes, though the details differ).
But enzyme active sites are complex 3D structures and the specific details of the active site, which will determine whether a specific molecule can bind to it, are far from simple. Sometimes a very small changes in the molecule makes a huge change to the binding ability potentially making the drug either less or more effective. Multiple changes can have synergistic or contradictory effects. And it isn't easy to work out why as we often don't know the actual structure of the key binding site or sometimes even which enzyme is the target.
So pharma firms, when they find something that works a bit, tend to get their chemists to make a wide range of slightly modified molecules and test them all in vivo in the hope some will be better.
It isn't simple and there are often no clear reasons why one molecule works better than another molecule. The variety of 3D structures (of drugs or binding sites) is very large and, especially when the target site is poorly understood, impossible to predict what modifications will work.
And these two considerations ignore another important constraint: whether a specific molecule will cause other side-effects by interaction with some other target.
Given all the constraints, it is not a surprise that it is hard to predict whether a molecular modification will have an effect that is good or bad. and you probably should not expect clear patters in most cases.