Category
AI & Machine Learning
Industry Insights & Trends

Where drug discovery loses time, and how to win it back

October 8, 2026
Pharmaceutical pills next to a vial of medicine

Why better structural data management is key to accelerating early-stage drug discovery

By Dr. Neil Taylor

Bringing a new medicine to market still takes more than a decade and costs billions of dollars. Most of that time is spent in clinical trials, but the earliest stages of drug discovery set the odds for everything that follows. A weak starting molecule, or a design decision made on incomplete data, can cost years downstream.

The industry's attention is now on AI in drug discovery. That attention is justified, but it often skips a less glamorous question. Before any model can help, can a research team actually find, trust and reuse the scientific data it already has?

The hidden cost of scattered structural data

Structure-based drug discovery (SBDD) depends on three-dimensional information: protein structures, the ligands bound to them, and the measured activity of each compound. A large pharmaceutical research organisation may hold tens of thousands of in-house crystal and cryo-electron microscopy (cryo-EM) structures. These sit alongside public data from the Protein Data Bank (PDB) and, increasingly, AI-predicted protein models.

In many organisations, that data is spread across project folders, individual workstations and legacy systems. Each structure is stored in a slightly different way. Chains are labelled inconsistently, binding sites are defined differently, and links to assay data live in separate databases.

The result is predictable. Scientists spend hours preparing protein structures before they can ask a scientific question. Work is repeated because nobody can tell whether a similar protein–ligand complex was solved three years ago on another programme. Knowledge leaves the building when people do.

Three approaches that are changing early-stage drug discovery

1. Treating structural data as a shared, curated asset

The most effective research groups now manage protein structure data the way they manage compound libraries. Every structure is processed consistently, annotated with its binding site and ligand, and searchable across all projects.

This turns a pile of files into an institutional memory that medicinal chemists, computational chemists and structural biologists can all query.

2. Linking protein structure to activity

A structure tells you how a molecule binds. An assay tells you how well.

Bringing the two together lets teams ask molecular design questions directly, such as which interactions are shared by the most potent compounds in a series. That is where structure-based drug design moves from hindsight to foresight.

3. Using AI where the data is ready for it

Machine learning models for protein structure prediction, molecular docking and binding affinity prediction are improving quickly. Their output is only as reliable as the data used to train, validate and interpret them.

Organisations with clean, consistent, well-annotated structural data are the ones best positioned to get real value from AI in drug discovery. Others are discovering that a model cannot fix an input problem.

What good pharmaceutical software solutions look like

The right drug discovery software is not one more silo. When research teams evaluate pharmaceutical software solutions, including specialist pharma software solutions for structural data management, a few questions separate tools that save time from tools that add work:

  • Does it standardise automatically? Protein structures should be processed consistently every time, without manual clean-up.
  • Can scientists search by science, not by filename? Look for the ability to search by protein, ligand, binding site and molecular interaction.
  • Does it connect to the rest of the drug discovery pipeline? Activity data, molecular modelling tools and electronic lab notebooks should link in, not require data to be re-entered.
  • Does it scale securely? Proprietary protein structures are among a company's most valuable intellectual property. Deployment and access control matter, whether on-premises or in the company's own cloud.
  • Will people actually use it? An interface that suits both expert crystallographers and bench chemists is what turns a protein structure database into a daily research tool.

Efficiency is a scientific advantage

It is tempting to see pharmaceutical research software as back-office infrastructure. In practice, the time a team saves on structural data preparation is time returned to molecular design and scientific decision-making.

A medicinal chemist who can see every relevant protein–ligand complex in seconds can make better-informed choices. A drug discovery programme that builds on past structures rather than rediscovering them can move faster towards a quality candidate.

Innovation in drug discovery is often framed as a single breakthrough. More often, it is the steady removal of friction from thousands of small decisions.

About DesertSci

For more than 25 years, DesertSci has built scientific software for structure-based drug discovery. Its flagship platform, Proasis, is used by more than 60% of the top 25 global pharmaceutical companies to manage, analyse and share protein structure data, with a 98% customer retention rate.

Learn more at desertsci.com and follow Dr Neil Taylor on LinkedIn for more insights.