---
license: mit
---

<center> <h1> The AIntibody Challenge </h1> </center>
<h2> Understanding and Advancing the Applications of AI to Antibody Discovery. </h2>

[AIntibody](https://www.aintibody.org/) is a series of benchmarking exercises that aims to advance antibody discovery by assessing the ability of unique AI algorithms to enhance or expedite the process. As the exercises—or ‘competitions’—progress, each challenge will become progressively more difficult.

A peer-reviewed manuscript for the exercises will provide the antibody community with the findings, highlighting the most successful approaches as well as areas that need the most improvement.

The AIntibody series is being announced to the antibody scientific community via publication in [Nature Biotech](https://www.nature.com/articles/s41587-024-02469-9).

## [Challenge 2](https://www.aintibody.org/the-challenges/#chal2): In Silico Affinity Rank Prediction

Given the NGS datasets of an HCDR3 clustered selection output, generated from the library described in [Teixeira et al. 2022](https://pubmed.ncbi.nlm.nih.gov/36068722/), identify those sequences within the existing NGS dataset that encode the highest affinity antibodies (that have not already had their affinities determined) in three HCDR3 clusters (27F, 28F and 47F).


<table>
  <tr>
    <td> <b>field/ column</b> </td>
    <td> <b>Description [TYPE]</b> </td>
    <td> <b># or range of non-redundant values</b> </td>
  </tr>  
  <tr>
    <td>characterized </td>
    <td>TRUE if characterized by SPR [BOOL] </td>
    <td>
      <table style="width: 100%;">
       <tr>
         <td>
            <ul>
                <li>#VL+VH: 142</li>
                <li>#L3+H3: 74</li>
                <li>#H3: 55</li>
                <li>#H3 clusters: 40</li>
            </ul>
         </td>
          <td>
            <ul>
                <li>#VL+VH: 31,917</li>
                <li>#L3+H3: 4362</li>
                <li>#H3: 754</li>
                <li>#H3 clusters: 300<</li>
            </ul>
         </td>
        </tr>
      </table>
    </td>
  </tr>
  <tr>
    <td>lsa_bin</td>
    <td>experimentally determined bin group [INTEGER}</td>
    <td>values = [1,2,5, NA]</td>
  </tr>
  <tr>
    <td>cluster_cdr3_heavy</td>
    <td>unique identifier (eg: 57F) [STRING]</td>
    <td>300</td>
  </tr>
  <tr>
    <td>affinity</td>
    <td>LSA affinity in molarity (M) [FLOAT]</td>
    <td>
        <ul>
            <li>2.7x10<sup>-11</sup> - 2.1x10<sup>-3</sup> M</li>
            <li>NA = uncharacterized</li>
            <li>1.0×10<sup>-6</sup> M = characterized, weak affinity</li>
        </ul>
    </td>
  </tr>
  <tr>
    <td>on-rate</td>
    <td>LSA on-rate in M<sup>-1</sup>s<sup>-1</sup> [FLOAT]</td>
    <td>
        <ul>
            <li>1.2×10<sup>2</sup> – 4.5×10<sup>5</sup> M<sup>-1</sup>s<sup>-1</sup></li>
            <li>NA = uncharacterized OR characterized, slow on-rate</li>
        </ul>
    </td>
  </tr>
  <tr>
    <td>off-rate</td>
    <td>LSA on-rate in s<sup>-1</sup> [FLOAT]</td>
    <td>
        <ul>
            <li>1.0×10<sup>-5</sup> – 0.6×10<sup>-2</sup> s<sup>-1</sup></li>
            <li>NA = uncharacterized OR characterized, fast off-rate</li>
        </ul>
    </td>
  </tr>
  <tr>
    <td>sequence_aa_light</td>
    <td>amino acid [STRING]</td>
    <td>102aa to 120aa</td>
  </tr>
  <tr>
    <td>sequence_aa_heavy</td>
    <td>amino acid [STRING]</td>
    <td>113aa to 133aa</td>
  </tr>
  <tr>
    <td>cdr1_aa_heavy [IMGT]</td>
    <td>amino acid [STRING]</td>
    <td>7aa to 14aa</td>
  </tr>
  <tr>
    <td>cdr2_aa_heavy [IMGT]</td>
    <td>amino acid [STRING]</td>
    <td>6aa to 9aa</td>
  </tr>
  <tr>
    <td>cdr3_aa_heavy [IMGT]</td>
    <td>amino acid [STRING]</td>
    <td>6aa to 26aa</td>
  </tr>
  <tr>
    <td>cdr1_aa_light [IMGT]</td>
    <td>amino acid [STRING]</td>
    <td>6aa to 12aa</td>
  </tr>
  <tr>
    <td>cdr2_aa_light [KABAT]</td>
    <td>amino acid [STRING]</td>
    <td>6aa to 11aa</td>
  </tr>
  <tr>
    <td>cdr3_aa_light [IMGT]</td>
    <td>amino acid [STRING]</td>
    <td>5aa to 18aa</td>
  </tr>
  <tr>
    <td>relative_abundance_10nM</td>
    <td>relative abundance of the concatenated CDRs in the 10nM RBD sort round via NGS [FLOAT]</td>
    <td>0.0%-6.1%</td>
  </tr>
  <tr>
    <td>relative_abundance_1nM</td>
    <td>relative abundance of the concatenated CDRs in the 1nM RBD sort round via NGS [FLOAT]</td>
    <td>0.0%-9.5%</td>
  </tr>
</table>

Example of dataset format (exmaple ONLY; not actual data) for challenge 2 is available in the example_data folder.

### Frequently Asked Questions

1. Who can participate?
The competition is open to any individuals/groups/entities that want to help further antibody science. [Complete this form](https://www.aintibody.org/get-started/) to register.

2. Why a competition?
AI is certainly a disruptive innovation in many areas of life sciences. The role of AI in antibody discovery is widely touted but has yet to be proven. By collating key antibody scientists and informaticists, we aim to better understand the landscape and help expedite the development and evolution of applied AI.

3. When does it start?
Participants can register beginning November 4 through mid-January. Registration may end sooner, as participation is capped at **1200 entries**. Datasets for each challenge will be distributed on **January 23**, with participants having **14 days** to provide their sequence data.

4. Additional Questions
The AIntibody program is administered by [Specifica](https://www.specifica.bio/), an IQVIA business.

View the [Terms of Participation](https://www.aintibody.org/terms-of-participation/).

Email any questions to <u>AIntibody@iqvia.com</u>.

## How to Participate

Full details on how to participate in the challenges are available at [AIntibody Challenge - How to Participate](https://www.aintibody.org/how-to-participate/).