Merge pull request #367 from eweitz/refine-tissue-expression

Ensure top max-expressing tissue is shown, robustify curve shift

scripts/python/cache/tissue_cache.py

@@ -101,7 +101,7 @@ def get_summary(expressions):
         if s > 0:
             summary.append(round(s, 2))
 
-    if len(summary) >= 4:
+    if len(summary) >= 4 or (len(summary) > 0 and summary[-1] >= 100):
         num_bins = 10
         size = (max - min) / num_bins
         quantile_counts = [0] * num_bins
@@ -240,39 +240,81 @@ def summarize_top_tissues_by_gene(input_dir):
 
             trimmed_summary_by_gene_by_tissue = {}
             trimmed_summary_by_gene_by_tissue[gene] = {}
-            medians_by_tissue_index = []
+            medians_and_maxes_by_tissue_index = []
             j = 0
+            top_max = 0
+            top_max_tissue = ''
             for tissue in summary_by_gene_by_tissue[gene]:
                 summary = summary_by_gene_by_tissue[gene][tissue]
-                if len(summary) == 15: # 5 for box plot, 10 for KDE deciles
+                num_metrics = len(summary)
+                if num_metrics == 15: # 5 for box plot, 10 for KDE deciles
                     median = summary[2]
-                elif len(summary) == 14: # for minimum with a value of 0
+                    max = summary[4]
+                elif num_metrics == 14: # for minimum with a value of 0
                     median = summary[1]
+                    max = summary[3]
+                elif num_metrics == 13:
+                    # Occurs in "Breast - mammary tissue" in LALBA or CSN3
+                    median = 0
+                    max = summary[2]
+                elif num_metrics == 12:
+                    # Occurs in "Breast - mammary tissue" in LALBA or CSN3
+                    median = 0
+                    max = summary[1]
                 else:
                     median = 0
-                medians_by_tissue_index.append([tissue, median])
-            sorted_medians_by_tissue_index = sorted(
-                medians_by_tissue_index,
-                key=lambda x: x[1],
+                    max = 0
+                if max > top_max:
+                    top_max = max
+                    top_max_tissue = tissue
+                medians_and_maxes_by_tissue_index.append([tissue, median, max])
+            sorted_medians_and_maxes_by_tissue_index = sorted(
+                medians_and_maxes_by_tissue_index,
+                key=lambda x: (x[1], x[2]),
                 reverse=True
             )
-            top_tissues_by_median = [
-                tm[0] for tm in sorted_medians_by_tissue_index[:10]
-            ]
-            for tissue in top_tissues_by_median:
+            top_tissues_by_median_and_max = []
+            includes_top_max = False
+            for tm in sorted_medians_and_maxes_by_tissue_index[:10]:
+                top_tissues_by_median_and_max.append(tm[0])
+                if tm[0] == top_max_tissue:
+                    includes_top_max = True
+
+            if not includes_top_max and top_max != 0:
+                # Ensure a tissue with a low median but the highest max
+                # isn't excluded from the top 10, e.g. "Breast - mammary gland"
+                # for gene XDH
+                top_tissues_by_median_and_max[-1] = top_max_tissue
+
+            for tissue in top_tissues_by_median_and_max:
                 summary = summary_by_gene_by_tissue[gene][tissue]
-                if len(summary) < 4:
-                    # Skip summaries that where Q1 (or higher percentile) is 0
+                if len(summary) < 10:
                     continue
-                summary = ';'.join([str(s) for s in summary])
+                elif len(summary) == 13:
+                    max = summary[2]
+                    if max < 100:
+                        # Skip summaries where Q1 (or higher percentile) is 0
+                        # and max is less than 100 TPM (e.g. to not skip
+                        # "Breast - mammary tissue" in LALBA or CSN3)
+                        continue
+                summary_list = []
+                for s in summary:
+                    if s == 0:
+                        # Delate 0-integer to empty string
+                        summary_list.append('')
+                    elif s < 1:
+                        # Truncate e.g. 0.1234 to .1234
+                        summary_list.append(str(s)[1:])
+                    else:
+                        summary_list.append(str(s))
+                summary = ';'.join(summary_list)
                 if summary == '':
                     # Observed for e.g. MEF2AP1, a pseudogene
                     continue
                 tissue_index = tissues_by_index_unique.index(tissue)
                 tissue_and_summary = f'{tissue_index};{summary}'
                 output_row.append(tissue_and_summary)
 
-
             if i % 500 == 0:
                 print(f'Last tissue summary for gene {i} {gene}:')
                 print(tissue_and_summary)

src/js/init/caches/tissue-cache-worker.js

@@ -44,14 +44,21 @@ async function getTissueExpressions(gene, ideo) {
   const tissueExpressions = [];
   const rawExpressions = geneDataLine.split('\t').slice(1);
   for (let i = 0; i < rawExpressions.length; i++) {
-    const rawValues = rawExpressions[i].split(';');
+    const rawValues = rawExpressions[i].split(';').map(
+      v => v === '' ? 0 : v // inflate empty string to 0-integer
+    );
     const numValues = rawValues.length;
     if (numValues === 15) {
       rawValues.splice(1, 0, 0); // Insert number 0 at position 1
     } else if (numValues === 14) {
       // Min. and Q1 are 0
       rawValues.splice(1, 0, 0);
       rawValues.splice(1, 0, 0);
+    } else if (numValues === 13) {
+      // Min., Q1, and median are 0
+      rawValues.splice(1, 0, 0);
+      rawValues.splice(1, 0, 0);
+      rawValues.splice(1, 0, 0);
     }
     const tissueId = rawValues[0];
     const boxMetrics = rawValues.slice(1, 6);

src/js/kit/protein-color.js

@@ -229,7 +229,9 @@ export function getColors(domainType) {
     domainType.includes('KA1') || // e.g. MARK2
     domainType === 'V(D)J recombination-activating protein 1' || // e.g. RAG1
     domainType.toLowerCase().includes('opiod') || // e.g. PDYN
-    domainType === 'Corticotropin-releasing factor' // e.g. CRH
+    domainType === 'Corticotropin-releasing factor' || // e.g. CRH
+    domainType.includes('2Fe-2S ferredoxin') || // e.g. XDH
+    domainType.includes('acidic domain') // e.g. SYNCRIP
   ) {
     return [redderFaintRed, redderFaintRedLine];
   } else if (
@@ -239,7 +241,8 @@ export function getColors(domainType) {
     domainType.includes('MutS, clamp') ||
     domainType.includes('S5 domain 2-like') || // e.g. MLH1 in ACMG
     domainType.endsWith('CC1/2') || // e.g. PRKDC
-    domainType.includes('MUN') // e.g. UNC13C
+    domainType.includes('MUN') || // e.g. UNC13C
+    domainType.includes('second molybdopterin') // e.g. XDH
   ) {
     return [darkGreen, darkGreenLine];
   } else if (
@@ -313,13 +316,15 @@ export function getColors(domainType) {
     domainType.includes('MG1') ||
     domainType === 'Homocysteine-binding domain' ||
     domainType.startsWith('Acyl-CoA') && domainType.endsWith('N-terminal') ||
-    domainType === 'Clathrin light chain'
+    domainType === 'Clathrin light chain' ||
+    domainType === 'Hexokinase, N-terminal' // e.g. HK2
   ) {
     return [lightBlue, lightBlueLine];
   } else if (
     // Larger binding regions and miscellaneous
     domainType.includes('zinc-binding') ||
     domainType.includes('DNA-binding') ||
+    domainType === 'RUNT domain' || // a DNA-binding / PPI domain, e.g. RUNX1
     domainType === 'G protein-coupled receptor, rhodopsin-like' ||
     domainType.includes('CXC domain') ||
     domainType.includes('Homeobox domain') ||
@@ -446,7 +451,8 @@ export function getColors(domainType) {
     domainType.includes('multifunctional domain') ||
     domainType.includes('MutS, core') ||
     domainType === 'RAP domain' || // RNA-binding, e.g. FASTK
-    domainType.endsWith('CC5') // e.g. PRKDC
+    domainType.endsWith('CC5') || // e.g. PRKDC
+    domainType.includes('first molybdopterin') // e.g. XDH
   ) {
     return [seafoam, seafoamLine];
   } else if (
@@ -639,7 +645,8 @@ export function getColors(domainType) {
     domainType === 'Cobalamin (vitamin B12)-binding domain' ||
     domainType === 'Laminin domain II' ||
     domainType === 'Troponin I residues 1-32' || // e.g. TNNI3 in ACMG
-    domainType === 'KI67R' // KI67 / Chmadrin repeat
+    domainType === 'KI67R' || // KI67 / Chmadrin repeat
+    domainType.includes('FAD-binding')
   ) {
     return [orange, orangeLines];
   } else if (
@@ -777,7 +784,9 @@ export function getColors(domainType) {
     domainType === 'GPCR, family 2, secretin-like' ||
     domainType === 'GPCR, family 3, nine cysteines domain' ||
     domainType === 'G-protein coupled receptor' ||
-    domainType.toLowerCase().includes('orexin') // e.g. HCRTR2
+    domainType.toLowerCase().includes('orexin') || // e.g. HCRTR2
+
+    domainType === '[2Fe-2S]-binding'
   ) {
     return [darkOrange, darkOrangeLines];
   } else if (
@@ -796,7 +805,6 @@ export function getColors(domainType) {
   } else if (
     domainType.includes('inhibit') ||
     domainType.includes('central') ||
-    domainType === '[2Fe-2S]-binding' ||
     domainType.endsWith('tail') ||
     domainType.endsWith('helical domain') ||
     domainType.endsWith('helical domain HD2') ||
@@ -826,7 +834,6 @@ export function getColors(domainType) {
     domainType === 'Paired domain' || // found in eukaryotic transcription regulatory proteins involved in embryogenesis
     domainType === 'JmjC domain' ||
     domainType === 'BRK domain' ||
-    domainType.includes('FAD-binding') ||
     domainType.includes('MG3') ||
     domainType.toLowerCase().includes('polycomb') ||
     domainType.toLowerCase().includes('metallopeptidase') ||

src/js/kit/tissue.js

@@ -665,14 +665,14 @@ function focusMiniCurve(traceDom, ideo, reset=false) {
   const refTissue = reset ? null : traceDom.getAttribute('data-tissue');
 
   const numTissues = !ideo.showTissuesMore ? 10 : 3;
-  let tissueExpressions =
-    ideo.tissueExpressionsByGene[gene].slice(0, numTissues);
+  let tissueExpressions = ideo.tissueExpressionsByGene[gene];
 
   const maxPx = MINI_CURVE_WIDTH;
   const relative = true;
   const leftPx = 0;
   tissueExpressions =
-    setPxOffset(tissueExpressions, maxPx, relative, leftPx, refTissue);
+    setPxOffset(tissueExpressions, maxPx, relative, leftPx, refTissue)
+      .slice(0, numTissues);
 
   const height = MINI_CURVE_HEIGHT;
   tissueExpressions.forEach((teObject, i) => {