Target identification & indication discovery

Use K-Pro to identify and prioritize therapeutic targets, and to characterize the patient subgroups in which they're most relevant.

• Prioritize targets: rank therapeutic targets and gene candidates using statistical evidence from patient cohorts (TCGA, and MOSAIC in K-Pro), normal-tissue references (GTEx), proteomics (CPTAC), and curated biological knowledge bases.
• Characterize patient subgroups: identify distinct populations based on molecular profiles and clinical outcomes.

The pages below walk through each task with a tested example prompt:

• Explore a cohort
• Compare groups
• Rank candidate targets

Worked example: prioritizing ADC targets

Here is how K-Pro approaches target prioritization for antibody-drug conjugates (ADCs) or, similarly. A similar workflow applies to radioligand therapies (RLTs) for instance, except that protein localization on the cell membrane is not required.

1

You ask a question

A researcher prompts K-Pro: "Identify and prioritize top ADC targets for bladder cancer (BLCA)."

2

K-Pro reasons about the constraints

K-Pro translates the question into biological constraints and the analyses needed to test them. A viable ADC target must be expressed on the cell surface, be highly prevalent in BLCA tumors, and have minimal (ideally zero) expression in essential healthy tissues, to keep a safe therapeutic window.

K-Pro also takes resistance into account. Heterogeneous expression can drive resistance, so good targets are expressed in a high proportion of cells or regions within each tumor. Ideally, non-expressing regions are still reached through the bystander effect.

3

K-Pro analyzes bulk, single-cell, and spatial data

K-Pro analyzes bulk transcriptomics (TCGA, MOSAIC), as well as single-cell and spatial transcriptomics (MOSAIC), and scores each candidate against these criteria.

Targets score highly when they are:

• strongly expressed in tumor tissue and absent from normal tissue (compared against GTEx normal tissue and Human Protein Atlas data);
• expressed either uniformly within the tumor, across single cells and spatial regions, or in a pattern that lets the bystander effect reach nearby non-expressing tumor cells.

4

K-Pro returns ranked targets with the evidence

K-Pro returns a ranking with per-target evidence, as both text and interactive plots, so you can see why each target ranks where it does.

5

Optional: add the competitive landscape

If you ask for it, K-Pro can also factor in the competitive landscape for each target, using clinical trial data and web search, so recommendations reflect both the biology and the development context.

Because this example is about expression-based modalities (ADCs), features typically used for small molecules, such as genetic association, dependency or oncogene addiction, and biochemical pathways, are not relevant here, and K-Pro does not apply them.

At the end, you can ask K-Pro to summarize the analyses, or, where reports are enabled, compile them into a report you can export as a PDF. Where skills are enabled for your organization, the whole sequence can also be saved as a skill. That way you can rerun the same workflow, for another indication for example, without repeating each prompt.