sldsc_postprocessing_pipeline: align sd_annot names to polyfun .results categories (covers --snp-list mode)

DESCRIPTION

@@ -77,3 +77,4 @@ LinkingTo:
 NeedsCompilation: yes
 VignetteBuilder: knitr
 Config/roxygen2/version: 8.0.0
+RoxygenNote: 7.3.3

NAMESPACE

@@ -20,6 +20,8 @@ export(coloc_wrapper)
 export(colocboost_analysis_pipeline)
 export(compute_LD)
 export(compute_qtl_enrichment)
+export(compute_sldsc_M_ref)
+export(compute_sldsc_annot_sd)
 export(ctwas_bimfile_loader)
 export(dentist)
 export(dentist_single_window)
@@ -49,6 +51,7 @@ export(glmnet_weights)
 export(harmonize_gwas)
 export(harmonize_twas)
 export(invert_minmax_scaling)
+export(is_binary_sldsc_annot)
 export(l0learn_weights)
 export(lasso_weights)
 export(lassosum_rss)
@@ -79,6 +82,7 @@ export(match_ref_panel)
 export(mcp_weights)
 export(merge_mash_data)
 export(merge_sumstats_matrices)
+export(meta_sldsc_random)
 export(mr_analysis)
 export(mr_ash_rss_weights)
 export(mr_format)
@@ -97,14 +101,17 @@ export(prs_cs)
 export(prs_cs_weights)
 export(raiss)
 export(read_afreq)
+export(read_sldsc_trait)
 export(region_to_df)
 export(rss_analysis_pipeline)
 export(rss_basic_qc)
 export(scad_weights)
 export(sdpr)
 export(sdpr_weights)
 export(slalom)
+export(sldsc_postprocessing_pipeline)
 export(standardise_sumstats_columns)
+export(standardize_sldsc_trait)
 export(summary_stats_qc)
 export(susie_ash_weights)
 export(susie_inf_weights)
@@ -138,6 +145,7 @@ importFrom(IRanges,start)
 importFrom(S4Vectors,queryHits)
 importFrom(S4Vectors,subjectHits)
 importFrom(coloc,coloc.bf_bf)
+importFrom(data.table,fread)
 importFrom(doFuture,registerDoFuture)
 importFrom(dplyr,across)
 importFrom(dplyr,all_of)
@@ -198,6 +206,7 @@ importFrom(stats,optim)
 importFrom(stats,pchisq)
 importFrom(stats,pnorm)
 importFrom(stats,predict)
+importFrom(stats,qnorm)
 importFrom(stats,sd)
 importFrom(stats,setNames)
 importFrom(stats,var)

R/sldsc_wrapper.R

@@ -546,6 +546,11 @@ sldsc_postprocessing_pipeline <- function(trait_single_prefixes,
   message("[sldsc] Detecting binary vs continuous annotations...")
   is_binary_full <- is_binary_sldsc_annot(target_anno_dir = target_anno_dir)
 
+  # Polyfun renames target columns to `<col>_0` (file_idx=0 in --ref-ld-chr);
+  # mirror that suffix so intersect() with pivot_run$categories matches.
+  names(sd_annot_full)  <- paste0(names(sd_annot_full),  "_0")
+  names(is_binary_full) <- paste0(names(is_binary_full), "_0")
+
   # Auto-detect target categories from a representative run.
   if (is.null(target_categories)) {
     pivot_run <- NULL
@@ -564,6 +569,33 @@ sldsc_postprocessing_pipeline <- function(trait_single_prefixes,
     if (is.null(pivot_run))
       stop("sldsc_postprocessing_pipeline: cannot auto-detect target_categories.")
     target_categories <- intersect(pivot_run$categories, names(sd_annot_full))
+    # Fallback when .annot.gz names + "_0" do not match polyfun's .results
+    # Category. Happens when the pipeline ran with --snp-list, since
+    # ldsc.py --l2 hardcodes the LD score column to "L2" (single annot) or
+    # "<annot>L2" (joint), instead of preserving the .annot.gz names.
+    # Trust polyfun's invariant that target categories occupy the first
+    # length(sd_annot_full) rows of .results, then rename positionally.
+    if (length(target_categories) == 0L) {
+      n_target   <- length(sd_annot_full)
+      n_baseline <- length(pivot_run$categories) - n_target
+      old_names  <- names(sd_annot_full)
+      target_categories <- pivot_run$categories[seq_len(n_target)]
+      names(sd_annot_full)  <- target_categories
+      names(is_binary_full) <- target_categories
+      baseline_preview <- if (n_baseline > 0L)
+        paste(utils::head(pivot_run$categories[-seq_len(n_target)], 3), collapse = ", ")
+      else "(none)"
+      message(sprintf(paste0(
+        "[sldsc] sd_annot/is_binary names did not match polyfun .results categories;\n",
+        "        falling back to positional rename (target = first %d rows of .results)\n",
+        "        target  (%d): %s -> %s\n",
+        "        baseline (%d): %s%s"),
+        n_target,
+        n_target, paste(old_names, collapse = ", "),
+                  paste(target_categories, collapse = ", "),
+        n_baseline, baseline_preview,
+        if (n_baseline > 3L) ", ..." else ""))
+    }
     message(sprintf("[sldsc] Auto-detected %d target categories", length(target_categories)))
   }