When researchers are looking to start developing a drug, what characteristics do they look for in the potential proteins (assuming they already have good quality structural models)?
The specific characteristics sought after will vary widely, but that being said, there are a few key features many groups look for. I've talked to and heard a number of individuals at a number of different companies, and there are a lot of different techniques and pathways looked for. In general, however, the single, most important factor is value. If the drug won't make money, it's almost never considered "promising." Maybe that's a sad fact of the way the world works, but it's a fact. The average drug costs around $1 billion (USD) to bring to market so the successful ones damn well better recoup that (Orphan drugs and companies actively pursuing them often have an easier time of this, as the FDA is understandably less restrictive and fast-tracks their progress).
That's not actually so simple. It's a large, crowded market out there, and there is a lot of intellectual property floating around. A wise businesswoman told me that her company focuses on the middle-ground. A target that is too promising is probably already owned by Genentech, so they aim for the ones that are below the attention of the giants but still quite profitable.
After that, it's mainly what you might expect. High avidity, long half-life coupled with low toxicity (more on this later), small size (if possible), and a realistic time frame. A month can mean millions of dollars, so working for a decade is not necessarily a possibility. Often, companies will use high-throughput screens of thousands, even millions of small molecules to look for an effect. The screening process is usually extremely rigorous, and many potential candidates are eliminated as not good enough. Sometimes they do some fancy stuff such as 3D modeling but it is definitely a field where brute force can still be successful.
Often those screens are used against primary cells to test for toxicity. In fact, throughout the entire process, toxicity is a huge factor. A full 20%-40% of drug attrition is due to safety and liability issues. It's a constant balance: nothing can risk the health of healthy, control volunteers, but sometimes more risk is acceptable. The first HIV/AIDS drugs were incredibly toxic, but that was okay because it was still better than the alternative. These days, the toxicity requirements for HIV/AIDS treatments such as HAART are stricter. Short-term antibiotics are allowed more risk since you only take them for five/seven days, whereas long-term medication for chronic illnesses are held to much higher standards.
Basically, loss of eyesight might be preferable to death, but it's unacceptable if treating the flu. The mantra I've been told is three things:
Toxicologists are conservative, willing to throw the baby out with the bathwater. In short, Fail Early.