diff --git a/doc/source/get_predicted_gain.rst b/doc/source/get_predicted_gain.rst
new file mode 100644
index 0000000000000000000000000000000000000000..9dea792471e07a2846adb7a73d92f8f0c14ea103
--- /dev/null
+++ b/doc/source/get_predicted_gain.rst
@@ -0,0 +1,39 @@
+Compute JASS power gain from the genetic architecture of traits
+===============================================================
+
+In a recent study :cite:`suzuki2023trait`, we explore how the genetic architecture of the set of traits (heritability, genetic covariance, heritability undetected by the univariate test, ...) can be predictive of statistical power gain of the multi-trait test.
+
+
+
+We implement an additional command line tool to give access our predictive model (the **jass predict-gain** command). 
+This command allows the to score swiftly a large number of traits combinations and to focus on set of traits the most promising for multi-trait testing.
+
+To work the inittable provided to the **jass predict-gain** command must contain the genetic covariance between traits. 
+
+
+.. code-block:: shell
+
+    jass predict-gain --inittable-path inittable_curated_111_traits_20-03-2024.hdf5 --combination_path ./combination_example.tsv --gain-path predicted_gain.tsv
+
+
+The second argument (--combination_path) is a path to a file containing the set of traits to be scored.
+
+.. csv-table:: Set of traits
+  :widths: 20, 70
+  :header-rows: 1
+
+    GRP1,z_GIANT_HIP z_GLG_HDL z_GLG_LDL z_MAGIC_2HGLU-ADJBMI
+    GRP2,z_SPIRO-UKB_FVC z_SPIRO-UKB_FEV1 z_TAGC_ASTHMA
+
+When executed the command will created a report at --gain-path
+
+.. csv-table:: Predicted gain
+    :header-rows: 1
+
+    traits,k,avg_distance_cor,mean_gencov,avg_Neff,avg_h2,avg_perc_h2_diff_region,log10_mean_gencov,log10_avg_distance_cor,gain
+    ['z_SPIRO-UKB_FVC'; 'z_SPIRO-UKB_FEV1'; 'z_TAGC_ASTHMA'],0.1,0.1731946683845993,0.0637,0.3843393026739591,0.2785193310634847,0.7976315890930669,0.8139196701681637,0.8013809378674498,0.06428524764535551
+    ['z_GIANT_HIP'; 'z_GLG_HDL'; 'z_GLG_LDL'; 'z_MAGIC_2HGLU-ADJBMI'],0.2,0.14899001074867035,0.01535,0.12076877719858631,0.22628198390356655,0.9055326131023057,0.6573854616675169,0.7879956172999502,-0.010766494024690904
+
+The last column provide the predicted gain ("the higher the more promising"). Note that extrapoling on new data might give lesser performances than reported in :cite:`suzuki2023trait`.
+
+.. bibliography:: reference.bib
\ No newline at end of file
diff --git a/doc/source/index.rst b/doc/source/index.rst
index 270f8c072ce9a715e642250b4daf5aa1b29772b3..87967f4a2da31ea34bc97b1035c41e2f138b0896 100644
--- a/doc/source/index.rst
+++ b/doc/source/index.rst
@@ -14,6 +14,7 @@ JASS documentation
    install
    data_import
    generating_joint_analysis
+   get_predicted_gain
    command_line_usage
    web_usage
    web_admin
diff --git a/doc/source/reference.bib b/doc/source/reference.bib
index 469a6903a53fa58d8e729b5e0d0e3b16c820968f..901b14a9691c75bac3e660837b9f99777a24a58a 100644
--- a/doc/source/reference.bib
+++ b/doc/source/reference.bib
@@ -20,6 +20,13 @@
   publisher={Oxford University Press}
 }
 
+@article{suzuki2023trait,
+  title={Trait selection strategy in multi-trait GWAS: Boosting SNPs discoverability},
+  author={Suzuki, Yuka and M{\'e}nager, Herv{\'e} and Brancotte, Bryan and Vernet, Rapha{\"e}l and Nerin, Cyril and Boetto, Christophe and Auvergne, Antoine and Linhard, Christophe and Torchet, Rachel and Lechat, Pierre and others},
+  journal={bioRxiv},
+  year={2023},
+  publisher={Cold Spring Harbor Laboratory Preprints}
+}
 
 
 @article{Pasaniuc2014,