Iteru Provides Proven Technology
Why we are different?
Up till now many, AI drug discovery solutions showed limited success because of low accuracy. Iteru puts accuracy at the forefront of its objectives. It provides measurable and verifiable analysis of accuracy at different stages of its AI analysis.
Quantification of Accuracy
Iteru puts accuracy at the forefront of its objectives. At each step of its AI analysis, including preprocessing, Iteru measures the accuracy of extracting biomedical entities. The two figures below show extracted biomedical entities before and after applying Iteru’s data cleansing and preprocessing algorithms. Iteru uses an algorithm to measure the number of entities in raw data before and after data cleansing. After cleansing there is pronounced increase in the number of entities. The proof of concept platform attained an encouraging 86% accuracy compared to our objective of 90%. Future improvements in text extraction and data cleansing will make it possible to attain more than 90%.
Biomedical data: before Cleansing
Biomedical data: after Cleansing
How the Proof of Concept is Used
The figure below illustrates the workflow for Iteru’s AI Platform. On the left side a scientist specifies the objective of analysis. The software extracts related data from the data lake, cleanses and labels it. It is preprocessed in a format amenable for processing by AI. AI extracts relationship between diseases and biomedical entities and relationship between biomedical entities themselves. The result can be displayed in a textual or tabular form. Alternatively, the result is displayed in 3-D. Using the display, a user can select a disease or an entity. The selected item is displayed in block letters and other entities are displayed around it based on proximity. In Figure 1. KRAS was selected. The scientist can select any entity or disease to investigate the association of the selection with others. Iteru will enhance the display by providing statistical analysis of entities’ spatial closeness and interaction between them. A scientist will be able to click on an entity and a popup will provide information about the association between the entity and disease or with other entities.
