Activity, sequences, pipeline — the data chain in pharma R&D is long, running from target to compound, from activity to clinical, and on to pipeline status, with each link having a different data format and a different processing approach. This page explains the data characteristics of each link, and exactly which sub-tasks we handle.
Activity concentrations are commonly expressed in many units — nM, µM, µg·mL⁻¹ and others — and even within a single paper they may be mixed; IC50, EC50, Ki and Kd have different meanings and are not interchangeable. More critical still are the assay conditions: with different cell lines, temperatures and incubation times, one value is not comparable with another.
So when structuring, value + unit + metric type + test subject + assay conditions must be carried together; drop any one of them and this data point is discounted downstream.
The efficacy endpoints, safety data and subject information in clinical trial reports need their fields standardized by indication and clinical phase, while at the same time satisfying compliance requirements. Here the boundary is not technical — it is about which fields may be retained.
Drug type, target/MOA, indication, development phase, pipeline status — all of these change over time. A snapshot extracted once is distorted within a few months, and what this kind of data needs is a sustainable update mechanism.
These links can be executed separately or combined into one complete pipeline: the structures, activities, sequences and pipeline status of the same set of compounds land in mutually linkable tables, making joint target–compound–indication analysis easier downstream.
Pull the activity data scattered across patents and literature, together with its assay conditions: test substances, activity metrics, value units, test subjects and methods. Handles mixed metrics and inconsistent units, delivering a SAR dataset with traceability information.
Unit normalizationRange valuesTable extractionNo. These quantities differ in meaning and in measurement principle and are not interchangeable. Unifying them into one "activity value" looks tidy, but in fact makes the downstream model learn a target that mixes several physical meanings. Metric type is a separate field in our schema.
By project agreement, processing takes place in a customer-designated environment or an isolated domain, with no data leaving the domain and no secondary use. Field design follows the principle of "retaining only the minimum information needed for analysis", and personally identifiable information does not enter the delivered dataset.
Because pipeline status changes over time: the development phase advances, indications expand, and projects may be terminated. A snapshot extracted once is distorted within a few months, and what this kind of data needs is a sustainable update mechanism, not a one-off delivery.
The original wording is kept and a separate normalized field is created and linked to an authoritative naming system. Aliases, former names and abbreviations of the same target across different documents are very common, and replacing them outright would lose the verifiability of the original text.
Send us a representative paper or data table and we can run a small trial annotation first, so you can see directly how field design and boundary handling work.