diff --git a/README.md b/README.md
index 17f5318..d1648f6 100644
--- a/README.md
+++ b/README.md
@@ -42,11 +42,10 @@ Crucially, **`DonorData`** ensures that genetic data and single-cell modalities
 
 **cellink** offers a streamlined suite of tools for the entire analysis workflow, organized by task:
 
-**Variant preprocessing & QC**
-- [Variant preprocessing & annotation](https://cellink-docs.readthedocs.io/en/latest/tutorials/explore_annotations.html): quality control, annotation (VCF export/import), and selection of genetic variants.
-
 **Association testing**
-- [Pseudobulk eQTL mapping](https://cellink-docs.readthedocs.io/en/latest/tutorials/pseudobulk_eqtl.html)
+- [eQTL analysis with jaxQTL or tensorQTL](https://cellink-docs.readthedocs.io/en/latest/tutorials/pseudobulk_eqtl_jaxqtl_tensorqtl.html)
+- [eQTL analysis with SAIGE-QTL](https://cellink-docs.readthedocs.io/en/latest/tutorials/single_cell_eqtl_saigeqtl.html)
+- [Naive pseudobulk eQTL mapping](https://cellink-docs.readthedocs.io/en/latest/tutorials/pseudobulk_eqtl.html)
 - [Rare variant association studies](https://cellink-docs.readthedocs.io/en/latest/tutorials/burden_testing.html)
 - [Clumping & pruning](https://cellink-docs.readthedocs.io/en/latest/tutorials/clumping_pruning.html)
 - [Colocalization analysis](https://cellink-docs.readthedocs.io/en/latest/tutorials/colocalization.html)
@@ -62,11 +61,10 @@ Crucially, **`DonorData`** ensures that genetic data and single-cell modalities
 **Deep learning & representation learning**
 - [LIVI: donor-level representation learning](https://cellink-docs.readthedocs.io/en/latest/tutorials/livi.html)
 - [Scooby: single-cell-resolution sequence-to-coverage modeling & variant scoring](https://cellink-docs.readthedocs.io/en/latest/tutorials/scooby.html)
-- [Built-in dataloaders for deep learning](https://cellink-docs.readthedocs.io/en/latest/tutorials/run_dataloader.html)
+- [Built-in dataloaders for deep learning](https://cellink-docs.readthedocs.io/en/latest/tutorials/run_dataloader.html), including Multiple Instance Learning (MIL) over per-donor cell bags
 
-**Interoperability & data integration**
-- [eQTL analysis with jaxQTL or tensorQTL](https://cellink-docs.readthedocs.io/en/latest/tutorials/pseudobulk_eqtl_jaxqtl_tensorqtl.html)
-- [eQTL analysis with SAIGE-QTL](https://cellink-docs.readthedocs.io/en/latest/tutorials/single_cell_eqtl_saigeqtl.html)
+**Other**
+- [Variant preprocessing & annotation](https://cellink-docs.readthedocs.io/en/latest/tutorials/explore_annotations.html): quality control, annotation (VCF export/import), and selection of genetic variants.
 - [Integrating `DonorData` with EHR data](https://cellink-docs.readthedocs.io/en/latest/tutorials/ehrdataset.html)
 
 ## Getting Started
diff --git a/tests/test_gene_pair_effects.py b/tests/test_gene_pair_effects.py
index d492e07..94fda3f 100644
--- a/tests/test_gene_pair_effects.py
+++ b/tests/test_gene_pair_effects.py
@@ -25,16 +25,16 @@ def test_compare_gene_pair_effects_keeps_only_celltypes_with_both_genes(tmp_path
         [
             (GENE_A, VARIANT, 0.3, 0.05, 1e-8),
             (GENE_B, VARIANT, -0.25, 0.06, 1e-6),
-            (GENE_A, "1:200:A:G", 0.1, 0.05, 0.2),  # different variant: irrelevant
+            (GENE_A, "1:200:A:G", 0.1, 0.05, 0.2), 
         ],
     )
-    # only GENE_A has a row at VARIANT here; celltype must be dropped entirely
+    
     _write_celltype_parquet(tmp_path, "ukb_european", "T_CD4_naive", [(GENE_A, VARIANT, 0.1, 0.02, 0.01)])
 
     res = compare_gene_pair_effects(tmp_path, VARIANT, GENE_A, GENE_B, cohort="ukb_european")
 
     assert list(res["celltype"].unique()) == ["NK_CD16"]
-    assert list(res["gene"]) == [GENE_A, GENE_B]  # gene_a first, then gene_b, per celltype
+    assert list(res["gene"]) == [GENE_A, GENE_B]  
     assert list(res.columns) == ["gene", "variant_id", "beta", "se", "pval", "celltype"]
 
 
diff --git a/tests/test_tl_external_dense_trans_scan.py b/tests/test_tl_external_dense_trans_scan.py
index 6f07816..6449ad2 100644
--- a/tests/test_tl_external_dense_trans_scan.py
+++ b/tests/test_tl_external_dense_trans_scan.py
@@ -39,7 +39,7 @@ def test_run_dense_trans_scan_ranks_all_genes(dd, far_variant_id):
     assert list(df["rank_by_pval"]) == list(range(1, len(df) + 1))
     assert df["pval"].is_monotonic_increasing
     assert {"chrom", "start", "end"}.issubset(df.columns)
-    # a genuinely unlinked (far-away) variant shouldn't lose any gene to the cis-window filter
+    
     assert len(df) == dd.C.n_vars