# Prioritize targets

Use **Targets** to rank a disease's drug targets on their evidence, and to read the literature on any gene.

Targets is not part of the Free plan.

## Start a new atlas

1. Open **Targets** and select **New atlas**.
2. Enter the **Disease**. Diseases in the shared catalog are suggested as you type.
3. Optionally enter **Target symbols** to include. A name Open Targets cannot resolve to one human gene is left out and reported in the run's activity.
4. **Optional details** holds **Cohort size** (300 by default, at most 350) and **Ontology ID**.
5. Select **Create project**, then confirm the run in **Start Marvin**.

The run takes the disease's most strongly associated targets from Open Targets, with their genetics, expression, clinical, tractability and safety data, and builds the table. It reads no papers. If you do not start it, the project waits under **My assessments** as **Not built yet**.

## Read the table

Each row is a target, with its rank, score and coverage, then one column per evidence channel. Select a channel heading to show its columns; hover a column heading for its definition and source.

- The score combines four pillars: genetics, literature, expression and clinical. A target with little known evidence is pulled toward the cohort's mean.
- **Coverage** is the known share of the evidence the score weighs.
- Hover a blank or zero cell to see what it means and how the score treats it.
- Literature columns count independent supporting groups in human genetics, patient samples, in vivo and in vitro studies, up to 2. A value shown with how many papers were read is a lower bound.
- Select any value to open the records behind it.

Search by symbol, synonym or name. **Filters** narrows the rows, by target class among others.

## Change the weights

Select **Weights**. Set each pillar's multiplier from 0 to 2, and switch each scored column on or off with a weight from 0 to 2. **Reset to defaults** restores the defaults, **Equal weights** sets every multiplier and weight to 1, and **Select all** and **Deselect all** switch every scored column. Context columns appear in the table but are never weighted.

Weights are saved in your browser for that table. If your browser cannot store them, the panel says so.

## Compare two measures

Open **Compare** and choose a **Horizontal axis** and a **Vertical axis**, any channel or column. **Swap axes** exchanges them. Dashed lines at each axis's median divide the plot into four labelled quadrants. Points are coloured by the therapeutic direction the target's interpretation calls; targets missing either value are left out and counted.

## Read a target

Select a target to open its summary beside the table, including why it was included. **Full profile** shows its interpretation, each statement opening the records it cites, every value with its evidence, its records by family, and links to GeneCards, Open Targets, Ensembl and UniProt.

- **Find papers** opens an Atlas search for the target and the disease.
- **Discuss** opens the project's chat with the target's dossier attached.

## Investigate a gene

**Investigate** reads the abstracts of up to 100 of a gene's papers in the disease, records what each one shows, and rebuilds the table. It appears in a target's summary and profile, and beside its literature values. To investigate a gene the table does not hold, type its name in the table's search: **Investigate** adds it once HGNC resolves the name to one gene, or with a warning when HGNC cannot be reached.

When Marvin is idle, **Start Marvin** confirms the one-iteration run first; while a run is active, the request is queued for it. Each run calls models on the account that pays for the project.

Once a gene is read, **Continue reading** reads its next 100 papers, while papers are left and some literature column is below 2 groups. A note says so instead when no readable papers were found for the gene, or when its records file is full.

## Explore and extend the shared catalog

**Explore** lists the diseases published to the shared catalog. Opening one shows its table, read-only.

**Extend into a new project** copies a release into a new project of your own, with any targets you add. Added targets stay private, and Marvin reads the abstracts of a target's papers only when you Investigate it.

Tables built before the current Targets engine still open and are ranked by the current score. Investigate is not offered on them until they are migrated, and Extend copies such a release unchanged.