@teselagen/bio-parsers 0.3.9 → 0.4.1

package/index.js

@@ -6170,7 +6170,9 @@ lodash.exports;
 })(lodash, lodash.exports);
 var lodashExports = lodash.exports;
 const _ = /* @__PURE__ */ getDefaultExportFromCjs(lodashExports);
+const protein_letters_withUandX = "ACDEFGHIKLMNPQRSTVWYUX";
 const ambiguous_dna_letters = "GATCRYWSMKHBVDN";
+const ambiguous_rna_letters = "GAUCRYWSMKHBVDN";
 const aminoAcidToDegenerateDnaMap = {
   "-": "---",
   ".": "...",
@@ -10775,7 +10777,7 @@ const proteinAlphabet = {
     hydrophobicity: 1.8,
     colorByFamily: "#00FFFF",
     color: "hsl(327.3, 100%, 69%)",
-    mass: 89.1
+    mass: 71.0779
   },
   R: {
     value: "R",
@@ -10784,7 +10786,7 @@ const proteinAlphabet = {
     hydrophobicity: -4.5,
     colorByFamily: "#FFC0CB",
     color: "hsl(258.1, 100%, 69%)",
-    mass: 174.2
+    mass: 156.18568
   },
   N: {
     value: "N",
@@ -10793,7 +10795,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.5,
     colorByFamily: "#D3D3D3",
     color: "hsl(268.9, 100%, 69%)",
-    mass: 132.1
+    mass: 114.10264
   },
   D: {
     value: "D",
@@ -10802,7 +10804,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.5,
     colorByFamily: "#EE82EE",
     color: "hsl(268.9, 100%, 69%)",
-    mass: 133.1
+    mass: 115.0874
   },
   C: {
     value: "C",
@@ -10811,7 +10813,7 @@ const proteinAlphabet = {
     hydrophobicity: 2.5,
     colorByFamily: "#FFFF00",
     color: "hsl(335.1, 100%, 69%)",
-    mass: 121.2
+    mass: 103.1429
   },
   E: {
     value: "E",
@@ -10820,7 +10822,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.5,
     colorByFamily: "#EE82EE",
     color: "hsl(268.9, 100%, 69%)",
-    mass: 147.1
+    mass: 129.11398
   },
   Q: {
     value: "Q",
@@ -10829,7 +10831,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.5,
     colorByFamily: "#D3D3D3",
     color: "hsl(268.9, 100%, 69%)",
-    mass: 146.2
+    mass: 128.12922
   },
   G: {
     value: "G",
@@ -10838,7 +10840,7 @@ const proteinAlphabet = {
     hydrophobicity: -0.4,
     colorByFamily: "#00FFFF",
     color: "hsl(303.1, 100%, 69%)",
-    mass: 75.1
+    mass: 57.05132
   },
   H: {
     value: "H",
@@ -10847,7 +10849,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.2,
     colorByFamily: "#FFC0CB",
     color: "hsl(272.2, 100%, 69%)",
-    mass: 155.2
+    mass: 137.13928
   },
   I: {
     value: "I",
@@ -10856,7 +10858,7 @@ const proteinAlphabet = {
     hydrophobicity: 4.5,
     colorByFamily: "#00FFFF",
     color: "hsl(356.9, 100%, 69%)",
-    mass: 131.2
+    mass: 113.15764
   },
   L: {
     value: "L",
@@ -10865,7 +10867,7 @@ const proteinAlphabet = {
     hydrophobicity: 3.8,
     colorByFamily: "#00FFFF",
     color: "hsl(349.4, 100%, 69%)",
-    mass: 131.2
+    mass: 113.15764
   },
   K: {
     value: "K",
@@ -10874,7 +10876,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.9,
     colorByFamily: "#FFC0CB",
     color: "hsl(264.7, 100%, 69%)",
-    mass: 146.2
+    mass: 128.17228
   },
   M: {
     value: "M",
@@ -10883,7 +10885,7 @@ const proteinAlphabet = {
     hydrophobicity: 1.9,
     colorByFamily: "#FFFF00",
     color: "hsl(328.5, 100%, 69%)",
-    mass: 149.2
+    mass: 131.19606
   },
   F: {
     value: "F",
@@ -10892,7 +10894,7 @@ const proteinAlphabet = {
     hydrophobicity: 2.8,
     colorByFamily: "#FFA500",
     color: "hsl(338.4, 100%, 69%)",
-    mass: 165.2
+    mass: 147.17386
   },
   P: {
     value: "P",
@@ -10901,7 +10903,7 @@ const proteinAlphabet = {
     hydrophobicity: -1.6,
     colorByFamily: "#00FFFF",
     color: "hsl(289.9, 100%, 69%)",
-    mass: 115.1
+    mass: 97.11518
   },
   S: {
     value: "S",
@@ -10910,7 +10912,7 @@ const proteinAlphabet = {
     hydrophobicity: -0.8,
     colorByFamily: "#90EE90",
     color: "hsl(298.6, 100%, 69%)",
-    mass: 105.1
+    mass: 87.0773
   },
   T: {
     value: "T",
@@ -10919,7 +10921,7 @@ const proteinAlphabet = {
     hydrophobicity: -0.7,
     colorByFamily: "#90EE90",
     color: "hsl(299.8, 100%, 69%)",
-    mass: 119.1
+    mass: 101.10388
   },
   U: {
     value: "U",
@@ -10927,7 +10929,7 @@ const proteinAlphabet = {
     threeLettersName: "Sec",
     colorByFamily: "#FF0000",
     color: "hsl(0, 100%, 69%)",
-    mass: 168.1
+    mass: 150.3079
   },
   W: {
     value: "W",
@@ -10936,7 +10938,7 @@ const proteinAlphabet = {
     hydrophobicity: -0.9,
     colorByFamily: "#FFA500",
     color: "hsl(297.6, 100%, 69%)",
-    mass: 204.2
+    mass: 186.2099
   },
   Y: {
     value: "Y",
@@ -10945,7 +10947,7 @@ const proteinAlphabet = {
     hydrophobicity: -1.3,
     colorByFamily: "#FFA500",
     color: "hsl(293.2, 100%, 69%)",
-    mass: 181.2
+    mass: 163.17326
   },
   V: {
     value: "V",
@@ -10954,7 +10956,7 @@ const proteinAlphabet = {
     hydrophobicity: 4.2,
     colorByFamily: "#00FFFF",
     color: "hsl(353.6, 100%, 69%)",
-    mass: 117.1
+    mass: 99.13106
   },
   "*": {
     value: "*",
@@ -11273,20 +11275,91 @@ const annotationTypes = [
   "primers",
   "guides"
 ];
-function filterSequenceString(sequenceString, additionalValidChars = "", charOverrides) {
-  if (sequenceString) {
-    return sequenceString.replace(
-      new RegExp(
-        `[^${charOverrides || `atgcyrswkmbvdhnu${additionalValidChars.split("").join("\\")}`}]`,
-        "gi"
-      ),
-      ""
+function filterSequenceString(sequenceString, {
+  additionalValidChars = "",
+  isOligo,
+  name,
+  isProtein,
+  isRna,
+  isMixedRnaAndDna,
+  includeStopCodon
+} = {}) {
+  const acceptedChars = getAcceptedChars({
+    isOligo,
+    isProtein,
+    isRna,
+    isMixedRnaAndDna,
+    includeStopCodon
+  });
+  const replaceChars = getReplaceChars({
+    isOligo,
+    isProtein,
+    isRna,
+    isMixedRnaAndDna
+  });
+  let sanitizedVal = "";
+  const invalidChars = [];
+  const chars = `${acceptedChars}${additionalValidChars.split("").join("\\")}`;
+  const warnings = [];
+  const replaceCount = {};
+  sequenceString.split("").forEach((letter) => {
+    const lowerLetter = letter.toLowerCase();
+    if (replaceChars && replaceChars[lowerLetter]) {
+      if (!replaceCount[lowerLetter]) {
+        replaceCount[lowerLetter] = 0;
+      }
+      replaceCount[lowerLetter]++;
+      const isUpper = lowerLetter !== letter;
+      sanitizedVal += isUpper ? replaceChars[lowerLetter].toUpperCase() : replaceChars[lowerLetter];
+    } else if (chars.includes(lowerLetter)) {
+      sanitizedVal += letter;
+    } else {
+      invalidChars.push(letter);
+    }
+  });
+  Object.keys(replaceCount).forEach((letter) => {
+    warnings.push(
+      `Replaced "${letter}" with "${replaceChars[letter]}"${replaceCount[letter] > 1 ? ` ${replaceCount[letter]} times` : ""}`
+    );
+  });
+  if (sequenceString.length !== sanitizedVal.length) {
+    warnings.push(
+      `${name ? `Sequence ${name}: ` : ""}Invalid character(s) detected and removed: ${invalidChars.slice(0, 100).join(", ")} `
     );
-  } else {
-    return sequenceString;
   }
+  if (typeof window !== "undefined" && window.toastr && warnings.length) {
+    warnings.forEach((warning) => {
+      window.toastr.warning(warning);
+    });
+  }
+  return [sanitizedVal, warnings];
 }
 __name(filterSequenceString, "filterSequenceString");
+function getAcceptedChars({
+  isOligo,
+  isProtein,
+  isRna,
+  isMixedRnaAndDna,
+  includeStopCodon
+} = {}) {
+  return isProtein ? `${protein_letters_withUandX.toLowerCase()}${includeStopCodon ? "*." : ""}}` : isOligo ? ambiguous_rna_letters.toLowerCase() + "t" : isRna ? ambiguous_rna_letters.toLowerCase() + "t" : isMixedRnaAndDna ? ambiguous_rna_letters.toLowerCase() + ambiguous_dna_letters.toLowerCase() : (
+    //just plain old dna
+    ambiguous_rna_letters.toLowerCase() + ambiguous_dna_letters.toLowerCase()
+  );
+}
+__name(getAcceptedChars, "getAcceptedChars");
+function getReplaceChars({
+  isOligo,
+  isProtein,
+  isRna,
+  isMixedRnaAndDna
+} = {}) {
+  return isProtein ? {} : isOligo ? {} : isRna ? { t: "u" } : isMixedRnaAndDna ? {} : (
+    //just plain old dna
+    {}
+  );
+}
+__name(getReplaceChars, "getReplaceChars");
 function tidyUpAnnotation(_annotation, {
   sequenceData = {},
   convertAnnotationsFromAAIndices,
@@ -11415,14 +11488,6 @@ function coerceLocation({
   }
 }
 __name(coerceLocation, "coerceLocation");
-function filterAminoAcidSequenceString(sequenceString, options) {
-  options = options || {};
-  if (options.includeStopCodon) {
-    return sequenceString == null ? void 0 : sequenceString.replace(/[^xtgalmfwkqespvicyhrndu.*]/gi, "");
-  }
-  return sequenceString == null ? void 0 : sequenceString.replace(/[^xtgalmfwkqespvicyhrndu]/gi, "");
-}
-__name(filterAminoAcidSequenceString, "filterAminoAcidSequenceString");
 function getDegenerateDnaStringFromAAString(aaString) {
   return aaString.split("").map((char) => aminoAcidToDegenerateDnaMap[char.toLowerCase()] || "nnn").join("");
 }
@@ -11434,11 +11499,10 @@ function tidyUpSequenceData(pSeqData, options = {}) {
     removeUnwantedChars,
     additionalValidChars,
     noTranslationData,
-    charOverrides,
     doNotProvideIdsForAnnotations,
-    proteinFilterOptions,
     noCdsTranslations,
-    convertAnnotationsFromAAIndices
+    convertAnnotationsFromAAIndices,
+    topLevelSeqData
   } = options;
   let seqData = lodashExports.cloneDeep(pSeqData);
   const response = {
@@ -11468,16 +11532,15 @@ function tidyUpSequenceData(pSeqData, options = {}) {
   }
   if (removeUnwantedChars) {
     if (seqData.isProtein) {
-      seqData.proteinSequence = filterAminoAcidSequenceString(
-        seqData.proteinSequence,
-        __spreadValues({ includeStopCodon: true }, proteinFilterOptions)
-      );
+      const [newSeq] = filterSequenceString(seqData.proteinSequence, __spreadValues({
+        includeStopCodon: true
+      }, topLevelSeqData || seqData));
+      seqData.proteinSequence = newSeq;
     } else {
-      seqData.sequence = filterSequenceString(
-        seqData.sequence,
-        `${additionalValidChars || ""}${seqData.isRna || seqData.isMixedRnaAndDna ? "u" : ""}`,
-        charOverrides
-      );
+      const [newSeq] = filterSequenceString(seqData.sequence, __spreadValues({
+        additionalValidChars
+      }, topLevelSeqData || seqData));
+      seqData.sequence = newSeq;
     }
   }
   if (seqData.isProtein) {
@@ -19312,7 +19375,8 @@ function validateSequence(sequence, options = {}) {
     inclusive1BasedEnd,
     additionalValidChars,
     allowOverflowAnnotations,
-    coerceFeatureTypes
+    coerceFeatureTypes,
+    includeStopCodon
   } = options;
   [
     "isDNA",
@@ -19362,7 +19426,6 @@ function validateSequence(sequence, options = {}) {
     response.messages.push("No sequence detected");
     sequence.sequence = "";
   }
-  let validChars;
   if (sequence.isProtein === void 0 && guessIfProtein) {
     sequence.isProtein = !guessIfSequenceIsDnaAndNotProtein(
       sequence.sequence,
@@ -19370,12 +19433,15 @@ function validateSequence(sequence, options = {}) {
     );
   }
   if (sequence.isProtein) {
-    validChars = filterAminoAcidSequenceString(sequence.sequence);
+    const [validChars, warnings] = filterSequenceString(sequence.sequence, {
+      name: sequence.name,
+      isProtein: true,
+      additionalValidChars,
+      includeStopCodon
+    });
     if (validChars !== sequence.sequence) {
       sequence.sequence = validChars;
-      response.messages.push(
-        "Import Error: Illegal character(s) detected and removed from amino acid sequence. Allowed characters are: xtgalmfwkqespvicyhrndu"
-      );
+      response.messages.push(...warnings);
     }
     sequence.type = "PROTEIN";
     sequence.isProtein = true;
@@ -19397,12 +19463,12 @@ function validateSequence(sequence, options = {}) {
     } else {
       sequence.type = "DNA";
     }
-    validChars = filterSequenceString(sequence.sequence, additionalValidChars);
+    const [validChars, warnings] = filterSequenceString(sequence.sequence, __spreadValues({
+      additionalValidChars
+    }, sequence));
     if (validChars !== sequence.sequence) {
       sequence.sequence = validChars;
-      response.messages.push(
-        "Import Error: Illegal character(s) detected and removed from sequence. Allowed characters are: atgcyrswkmbvdhn"
-      );
+      response.messages.push(...warnings);
     }
   }
   if (!sequence.size) {

package/index.mjs

@@ -6168,7 +6168,9 @@ lodash.exports;
 })(lodash, lodash.exports);
 var lodashExports = lodash.exports;
 const _ = /* @__PURE__ */ getDefaultExportFromCjs(lodashExports);
+const protein_letters_withUandX = "ACDEFGHIKLMNPQRSTVWYUX";
 const ambiguous_dna_letters = "GATCRYWSMKHBVDN";
+const ambiguous_rna_letters = "GAUCRYWSMKHBVDN";
 const aminoAcidToDegenerateDnaMap = {
   "-": "---",
   ".": "...",
@@ -10773,7 +10775,7 @@ const proteinAlphabet = {
     hydrophobicity: 1.8,
     colorByFamily: "#00FFFF",
     color: "hsl(327.3, 100%, 69%)",
-    mass: 89.1
+    mass: 71.0779
   },
   R: {
     value: "R",
@@ -10782,7 +10784,7 @@ const proteinAlphabet = {
     hydrophobicity: -4.5,
     colorByFamily: "#FFC0CB",
     color: "hsl(258.1, 100%, 69%)",
-    mass: 174.2
+    mass: 156.18568
   },
   N: {
     value: "N",
@@ -10791,7 +10793,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.5,
     colorByFamily: "#D3D3D3",
     color: "hsl(268.9, 100%, 69%)",
-    mass: 132.1
+    mass: 114.10264
   },
   D: {
     value: "D",
@@ -10800,7 +10802,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.5,
     colorByFamily: "#EE82EE",
     color: "hsl(268.9, 100%, 69%)",
-    mass: 133.1
+    mass: 115.0874
   },
   C: {
     value: "C",
@@ -10809,7 +10811,7 @@ const proteinAlphabet = {
     hydrophobicity: 2.5,
     colorByFamily: "#FFFF00",
     color: "hsl(335.1, 100%, 69%)",
-    mass: 121.2
+    mass: 103.1429
   },
   E: {
     value: "E",
@@ -10818,7 +10820,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.5,
     colorByFamily: "#EE82EE",
     color: "hsl(268.9, 100%, 69%)",
-    mass: 147.1
+    mass: 129.11398
   },
   Q: {
     value: "Q",
@@ -10827,7 +10829,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.5,
     colorByFamily: "#D3D3D3",
     color: "hsl(268.9, 100%, 69%)",
-    mass: 146.2
+    mass: 128.12922
   },
   G: {
     value: "G",
@@ -10836,7 +10838,7 @@ const proteinAlphabet = {
     hydrophobicity: -0.4,
     colorByFamily: "#00FFFF",
     color: "hsl(303.1, 100%, 69%)",
-    mass: 75.1
+    mass: 57.05132
   },
   H: {
     value: "H",
@@ -10845,7 +10847,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.2,
     colorByFamily: "#FFC0CB",
     color: "hsl(272.2, 100%, 69%)",
-    mass: 155.2
+    mass: 137.13928
   },
   I: {
     value: "I",
@@ -10854,7 +10856,7 @@ const proteinAlphabet = {
     hydrophobicity: 4.5,
     colorByFamily: "#00FFFF",
     color: "hsl(356.9, 100%, 69%)",
-    mass: 131.2
+    mass: 113.15764
   },
   L: {
     value: "L",
@@ -10863,7 +10865,7 @@ const proteinAlphabet = {
     hydrophobicity: 3.8,
     colorByFamily: "#00FFFF",
     color: "hsl(349.4, 100%, 69%)",
-    mass: 131.2
+    mass: 113.15764
   },
   K: {
     value: "K",
@@ -10872,7 +10874,7 @@ const proteinAlphabet = {
     hydrophobicity: -3.9,
     colorByFamily: "#FFC0CB",
     color: "hsl(264.7, 100%, 69%)",
-    mass: 146.2
+    mass: 128.17228
   },
   M: {
     value: "M",
@@ -10881,7 +10883,7 @@ const proteinAlphabet = {
     hydrophobicity: 1.9,
     colorByFamily: "#FFFF00",
     color: "hsl(328.5, 100%, 69%)",
-    mass: 149.2
+    mass: 131.19606
   },
   F: {
     value: "F",
@@ -10890,7 +10892,7 @@ const proteinAlphabet = {
     hydrophobicity: 2.8,
     colorByFamily: "#FFA500",
     color: "hsl(338.4, 100%, 69%)",
-    mass: 165.2
+    mass: 147.17386
   },
   P: {
     value: "P",
@@ -10899,7 +10901,7 @@ const proteinAlphabet = {
     hydrophobicity: -1.6,
     colorByFamily: "#00FFFF",
     color: "hsl(289.9, 100%, 69%)",
-    mass: 115.1
+    mass: 97.11518
   },
   S: {
     value: "S",
@@ -10908,7 +10910,7 @@ const proteinAlphabet = {
     hydrophobicity: -0.8,
     colorByFamily: "#90EE90",
     color: "hsl(298.6, 100%, 69%)",
-    mass: 105.1
+    mass: 87.0773
   },
   T: {
     value: "T",
@@ -10917,7 +10919,7 @@ const proteinAlphabet = {
     hydrophobicity: -0.7,
     colorByFamily: "#90EE90",
     color: "hsl(299.8, 100%, 69%)",
-    mass: 119.1
+    mass: 101.10388
   },
   U: {
     value: "U",
@@ -10925,7 +10927,7 @@ const proteinAlphabet = {
     threeLettersName: "Sec",
     colorByFamily: "#FF0000",
     color: "hsl(0, 100%, 69%)",
-    mass: 168.1
+    mass: 150.3079
   },
   W: {
     value: "W",
@@ -10934,7 +10936,7 @@ const proteinAlphabet = {
     hydrophobicity: -0.9,
     colorByFamily: "#FFA500",
     color: "hsl(297.6, 100%, 69%)",
-    mass: 204.2
+    mass: 186.2099
   },
   Y: {
     value: "Y",
@@ -10943,7 +10945,7 @@ const proteinAlphabet = {
     hydrophobicity: -1.3,
     colorByFamily: "#FFA500",
     color: "hsl(293.2, 100%, 69%)",
-    mass: 181.2
+    mass: 163.17326
   },
   V: {
     value: "V",
@@ -10952,7 +10954,7 @@ const proteinAlphabet = {
     hydrophobicity: 4.2,
     colorByFamily: "#00FFFF",
     color: "hsl(353.6, 100%, 69%)",
-    mass: 117.1
+    mass: 99.13106
   },
   "*": {
     value: "*",
@@ -11271,20 +11273,91 @@ const annotationTypes = [
   "primers",
   "guides"
 ];
-function filterSequenceString(sequenceString, additionalValidChars = "", charOverrides) {
-  if (sequenceString) {
-    return sequenceString.replace(
-      new RegExp(
-        `[^${charOverrides || `atgcyrswkmbvdhnu${additionalValidChars.split("").join("\\")}`}]`,
-        "gi"
-      ),
-      ""
+function filterSequenceString(sequenceString, {
+  additionalValidChars = "",
+  isOligo,
+  name,
+  isProtein,
+  isRna,
+  isMixedRnaAndDna,
+  includeStopCodon
+} = {}) {
+  const acceptedChars = getAcceptedChars({
+    isOligo,
+    isProtein,
+    isRna,
+    isMixedRnaAndDna,
+    includeStopCodon
+  });
+  const replaceChars = getReplaceChars({
+    isOligo,
+    isProtein,
+    isRna,
+    isMixedRnaAndDna
+  });
+  let sanitizedVal = "";
+  const invalidChars = [];
+  const chars = `${acceptedChars}${additionalValidChars.split("").join("\\")}`;
+  const warnings = [];
+  const replaceCount = {};
+  sequenceString.split("").forEach((letter) => {
+    const lowerLetter = letter.toLowerCase();
+    if (replaceChars && replaceChars[lowerLetter]) {
+      if (!replaceCount[lowerLetter]) {
+        replaceCount[lowerLetter] = 0;
+      }
+      replaceCount[lowerLetter]++;
+      const isUpper = lowerLetter !== letter;
+      sanitizedVal += isUpper ? replaceChars[lowerLetter].toUpperCase() : replaceChars[lowerLetter];
+    } else if (chars.includes(lowerLetter)) {
+      sanitizedVal += letter;
+    } else {
+      invalidChars.push(letter);
+    }
+  });
+  Object.keys(replaceCount).forEach((letter) => {
+    warnings.push(
+      `Replaced "${letter}" with "${replaceChars[letter]}"${replaceCount[letter] > 1 ? ` ${replaceCount[letter]} times` : ""}`
+    );
+  });
+  if (sequenceString.length !== sanitizedVal.length) {
+    warnings.push(
+      `${name ? `Sequence ${name}: ` : ""}Invalid character(s) detected and removed: ${invalidChars.slice(0, 100).join(", ")} `
     );
-  } else {
-    return sequenceString;
   }
+  if (typeof window !== "undefined" && window.toastr && warnings.length) {
+    warnings.forEach((warning) => {
+      window.toastr.warning(warning);
+    });
+  }
+  return [sanitizedVal, warnings];
 }
 __name(filterSequenceString, "filterSequenceString");
+function getAcceptedChars({
+  isOligo,
+  isProtein,
+  isRna,
+  isMixedRnaAndDna,
+  includeStopCodon
+} = {}) {
+  return isProtein ? `${protein_letters_withUandX.toLowerCase()}${includeStopCodon ? "*." : ""}}` : isOligo ? ambiguous_rna_letters.toLowerCase() + "t" : isRna ? ambiguous_rna_letters.toLowerCase() + "t" : isMixedRnaAndDna ? ambiguous_rna_letters.toLowerCase() + ambiguous_dna_letters.toLowerCase() : (
+    //just plain old dna
+    ambiguous_rna_letters.toLowerCase() + ambiguous_dna_letters.toLowerCase()
+  );
+}
+__name(getAcceptedChars, "getAcceptedChars");
+function getReplaceChars({
+  isOligo,
+  isProtein,
+  isRna,
+  isMixedRnaAndDna
+} = {}) {
+  return isProtein ? {} : isOligo ? {} : isRna ? { t: "u" } : isMixedRnaAndDna ? {} : (
+    //just plain old dna
+    {}
+  );
+}
+__name(getReplaceChars, "getReplaceChars");
 function tidyUpAnnotation(_annotation, {
   sequenceData = {},
   convertAnnotationsFromAAIndices,
@@ -11413,14 +11486,6 @@ function coerceLocation({
   }
 }
 __name(coerceLocation, "coerceLocation");
-function filterAminoAcidSequenceString(sequenceString, options) {
-  options = options || {};
-  if (options.includeStopCodon) {
-    return sequenceString == null ? void 0 : sequenceString.replace(/[^xtgalmfwkqespvicyhrndu.*]/gi, "");
-  }
-  return sequenceString == null ? void 0 : sequenceString.replace(/[^xtgalmfwkqespvicyhrndu]/gi, "");
-}
-__name(filterAminoAcidSequenceString, "filterAminoAcidSequenceString");
 function getDegenerateDnaStringFromAAString(aaString) {
   return aaString.split("").map((char) => aminoAcidToDegenerateDnaMap[char.toLowerCase()] || "nnn").join("");
 }
@@ -11432,11 +11497,10 @@ function tidyUpSequenceData(pSeqData, options = {}) {
     removeUnwantedChars,
     additionalValidChars,
     noTranslationData,
-    charOverrides,
     doNotProvideIdsForAnnotations,
-    proteinFilterOptions,
     noCdsTranslations,
-    convertAnnotationsFromAAIndices
+    convertAnnotationsFromAAIndices,
+    topLevelSeqData
   } = options;
   let seqData = lodashExports.cloneDeep(pSeqData);
   const response = {
@@ -11466,16 +11530,15 @@ function tidyUpSequenceData(pSeqData, options = {}) {
   }
   if (removeUnwantedChars) {
     if (seqData.isProtein) {
-      seqData.proteinSequence = filterAminoAcidSequenceString(
-        seqData.proteinSequence,
-        __spreadValues({ includeStopCodon: true }, proteinFilterOptions)
-      );
+      const [newSeq] = filterSequenceString(seqData.proteinSequence, __spreadValues({
+        includeStopCodon: true
+      }, topLevelSeqData || seqData));
+      seqData.proteinSequence = newSeq;
     } else {
-      seqData.sequence = filterSequenceString(
-        seqData.sequence,
-        `${additionalValidChars || ""}${seqData.isRna || seqData.isMixedRnaAndDna ? "u" : ""}`,
-        charOverrides
-      );
+      const [newSeq] = filterSequenceString(seqData.sequence, __spreadValues({
+        additionalValidChars
+      }, topLevelSeqData || seqData));
+      seqData.sequence = newSeq;
     }
   }
   if (seqData.isProtein) {
@@ -19310,7 +19373,8 @@ function validateSequence(sequence, options = {}) {
     inclusive1BasedEnd,
     additionalValidChars,
     allowOverflowAnnotations,
-    coerceFeatureTypes
+    coerceFeatureTypes,
+    includeStopCodon
   } = options;
   [
     "isDNA",
@@ -19360,7 +19424,6 @@ function validateSequence(sequence, options = {}) {
     response.messages.push("No sequence detected");
     sequence.sequence = "";
   }
-  let validChars;
   if (sequence.isProtein === void 0 && guessIfProtein) {
     sequence.isProtein = !guessIfSequenceIsDnaAndNotProtein(
       sequence.sequence,
@@ -19368,12 +19431,15 @@ function validateSequence(sequence, options = {}) {
     );
   }
   if (sequence.isProtein) {
-    validChars = filterAminoAcidSequenceString(sequence.sequence);
+    const [validChars, warnings] = filterSequenceString(sequence.sequence, {
+      name: sequence.name,
+      isProtein: true,
+      additionalValidChars,
+      includeStopCodon
+    });
     if (validChars !== sequence.sequence) {
       sequence.sequence = validChars;
-      response.messages.push(
-        "Import Error: Illegal character(s) detected and removed from amino acid sequence. Allowed characters are: xtgalmfwkqespvicyhrndu"
-      );
+      response.messages.push(...warnings);
     }
     sequence.type = "PROTEIN";
     sequence.isProtein = true;
@@ -19395,12 +19461,12 @@ function validateSequence(sequence, options = {}) {
     } else {
       sequence.type = "DNA";
     }
-    validChars = filterSequenceString(sequence.sequence, additionalValidChars);
+    const [validChars, warnings] = filterSequenceString(sequence.sequence, __spreadValues({
+      additionalValidChars
+    }, sequence));
     if (validChars !== sequence.sequence) {
       sequence.sequence = validChars;
-      response.messages.push(
-        "Import Error: Illegal character(s) detected and removed from sequence. Allowed characters are: atgcyrswkmbvdhn"
-      );
+      response.messages.push(...warnings);
     }
   }
   if (!sequence.size) {

package/index.umd.js

@@ -6172,7 +6172,9 @@ var __async = (__this, __arguments, generator) => {
   })(lodash, lodash.exports);
   var lodashExports = lodash.exports;
   const _ = /* @__PURE__ */ getDefaultExportFromCjs(lodashExports);
+  const protein_letters_withUandX = "ACDEFGHIKLMNPQRSTVWYUX";
   const ambiguous_dna_letters = "GATCRYWSMKHBVDN";
+  const ambiguous_rna_letters = "GAUCRYWSMKHBVDN";
   const aminoAcidToDegenerateDnaMap = {
     "-": "---",
     ".": "...",
@@ -10777,7 +10779,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: 1.8,
       colorByFamily: "#00FFFF",
       color: "hsl(327.3, 100%, 69%)",
-      mass: 89.1
+      mass: 71.0779
     },
     R: {
       value: "R",
@@ -10786,7 +10788,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -4.5,
       colorByFamily: "#FFC0CB",
       color: "hsl(258.1, 100%, 69%)",
-      mass: 174.2
+      mass: 156.18568
     },
     N: {
       value: "N",
@@ -10795,7 +10797,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -3.5,
       colorByFamily: "#D3D3D3",
       color: "hsl(268.9, 100%, 69%)",
-      mass: 132.1
+      mass: 114.10264
     },
     D: {
       value: "D",
@@ -10804,7 +10806,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -3.5,
       colorByFamily: "#EE82EE",
       color: "hsl(268.9, 100%, 69%)",
-      mass: 133.1
+      mass: 115.0874
     },
     C: {
       value: "C",
@@ -10813,7 +10815,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: 2.5,
       colorByFamily: "#FFFF00",
       color: "hsl(335.1, 100%, 69%)",
-      mass: 121.2
+      mass: 103.1429
     },
     E: {
       value: "E",
@@ -10822,7 +10824,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -3.5,
       colorByFamily: "#EE82EE",
       color: "hsl(268.9, 100%, 69%)",
-      mass: 147.1
+      mass: 129.11398
     },
     Q: {
       value: "Q",
@@ -10831,7 +10833,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -3.5,
       colorByFamily: "#D3D3D3",
       color: "hsl(268.9, 100%, 69%)",
-      mass: 146.2
+      mass: 128.12922
     },
     G: {
       value: "G",
@@ -10840,7 +10842,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -0.4,
       colorByFamily: "#00FFFF",
       color: "hsl(303.1, 100%, 69%)",
-      mass: 75.1
+      mass: 57.05132
     },
     H: {
       value: "H",
@@ -10849,7 +10851,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -3.2,
       colorByFamily: "#FFC0CB",
       color: "hsl(272.2, 100%, 69%)",
-      mass: 155.2
+      mass: 137.13928
     },
     I: {
       value: "I",
@@ -10858,7 +10860,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: 4.5,
       colorByFamily: "#00FFFF",
       color: "hsl(356.9, 100%, 69%)",
-      mass: 131.2
+      mass: 113.15764
     },
     L: {
       value: "L",
@@ -10867,7 +10869,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: 3.8,
       colorByFamily: "#00FFFF",
       color: "hsl(349.4, 100%, 69%)",
-      mass: 131.2
+      mass: 113.15764
     },
     K: {
       value: "K",
@@ -10876,7 +10878,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -3.9,
       colorByFamily: "#FFC0CB",
       color: "hsl(264.7, 100%, 69%)",
-      mass: 146.2
+      mass: 128.17228
     },
     M: {
       value: "M",
@@ -10885,7 +10887,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: 1.9,
       colorByFamily: "#FFFF00",
       color: "hsl(328.5, 100%, 69%)",
-      mass: 149.2
+      mass: 131.19606
     },
     F: {
       value: "F",
@@ -10894,7 +10896,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: 2.8,
       colorByFamily: "#FFA500",
       color: "hsl(338.4, 100%, 69%)",
-      mass: 165.2
+      mass: 147.17386
     },
     P: {
       value: "P",
@@ -10903,7 +10905,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -1.6,
       colorByFamily: "#00FFFF",
       color: "hsl(289.9, 100%, 69%)",
-      mass: 115.1
+      mass: 97.11518
     },
     S: {
       value: "S",
@@ -10912,7 +10914,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -0.8,
       colorByFamily: "#90EE90",
       color: "hsl(298.6, 100%, 69%)",
-      mass: 105.1
+      mass: 87.0773
     },
     T: {
       value: "T",
@@ -10921,7 +10923,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -0.7,
       colorByFamily: "#90EE90",
       color: "hsl(299.8, 100%, 69%)",
-      mass: 119.1
+      mass: 101.10388
     },
     U: {
       value: "U",
@@ -10929,7 +10931,7 @@ var __async = (__this, __arguments, generator) => {
       threeLettersName: "Sec",
       colorByFamily: "#FF0000",
       color: "hsl(0, 100%, 69%)",
-      mass: 168.1
+      mass: 150.3079
     },
     W: {
       value: "W",
@@ -10938,7 +10940,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -0.9,
       colorByFamily: "#FFA500",
       color: "hsl(297.6, 100%, 69%)",
-      mass: 204.2
+      mass: 186.2099
     },
     Y: {
       value: "Y",
@@ -10947,7 +10949,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: -1.3,
       colorByFamily: "#FFA500",
       color: "hsl(293.2, 100%, 69%)",
-      mass: 181.2
+      mass: 163.17326
     },
     V: {
       value: "V",
@@ -10956,7 +10958,7 @@ var __async = (__this, __arguments, generator) => {
       hydrophobicity: 4.2,
       colorByFamily: "#00FFFF",
       color: "hsl(353.6, 100%, 69%)",
-      mass: 117.1
+      mass: 99.13106
     },
     "*": {
       value: "*",
@@ -11275,20 +11277,91 @@ var __async = (__this, __arguments, generator) => {
     "primers",
     "guides"
   ];
-  function filterSequenceString(sequenceString, additionalValidChars = "", charOverrides) {
-    if (sequenceString) {
-      return sequenceString.replace(
-        new RegExp(
-          `[^${charOverrides || `atgcyrswkmbvdhnu${additionalValidChars.split("").join("\\")}`}]`,
-          "gi"
-        ),
-        ""
+  function filterSequenceString(sequenceString, {
+    additionalValidChars = "",
+    isOligo,
+    name: name2,
+    isProtein,
+    isRna,
+    isMixedRnaAndDna,
+    includeStopCodon
+  } = {}) {
+    const acceptedChars = getAcceptedChars({
+      isOligo,
+      isProtein,
+      isRna,
+      isMixedRnaAndDna,
+      includeStopCodon
+    });
+    const replaceChars = getReplaceChars({
+      isOligo,
+      isProtein,
+      isRna,
+      isMixedRnaAndDna
+    });
+    let sanitizedVal = "";
+    const invalidChars = [];
+    const chars = `${acceptedChars}${additionalValidChars.split("").join("\\")}`;
+    const warnings = [];
+    const replaceCount = {};
+    sequenceString.split("").forEach((letter) => {
+      const lowerLetter = letter.toLowerCase();
+      if (replaceChars && replaceChars[lowerLetter]) {
+        if (!replaceCount[lowerLetter]) {
+          replaceCount[lowerLetter] = 0;
+        }
+        replaceCount[lowerLetter]++;
+        const isUpper = lowerLetter !== letter;
+        sanitizedVal += isUpper ? replaceChars[lowerLetter].toUpperCase() : replaceChars[lowerLetter];
+      } else if (chars.includes(lowerLetter)) {
+        sanitizedVal += letter;
+      } else {
+        invalidChars.push(letter);
+      }
+    });
+    Object.keys(replaceCount).forEach((letter) => {
+      warnings.push(
+        `Replaced "${letter}" with "${replaceChars[letter]}"${replaceCount[letter] > 1 ? ` ${replaceCount[letter]} times` : ""}`
+      );
+    });
+    if (sequenceString.length !== sanitizedVal.length) {
+      warnings.push(
+        `${name2 ? `Sequence ${name2}: ` : ""}Invalid character(s) detected and removed: ${invalidChars.slice(0, 100).join(", ")} `
       );
-    } else {
-      return sequenceString;
     }
+    if (typeof window !== "undefined" && window.toastr && warnings.length) {
+      warnings.forEach((warning) => {
+        window.toastr.warning(warning);
+      });
+    }
+    return [sanitizedVal, warnings];
   }
   __name(filterSequenceString, "filterSequenceString");
+  function getAcceptedChars({
+    isOligo,
+    isProtein,
+    isRna,
+    isMixedRnaAndDna,
+    includeStopCodon
+  } = {}) {
+    return isProtein ? `${protein_letters_withUandX.toLowerCase()}${includeStopCodon ? "*." : ""}}` : isOligo ? ambiguous_rna_letters.toLowerCase() + "t" : isRna ? ambiguous_rna_letters.toLowerCase() + "t" : isMixedRnaAndDna ? ambiguous_rna_letters.toLowerCase() + ambiguous_dna_letters.toLowerCase() : (
+      //just plain old dna
+      ambiguous_rna_letters.toLowerCase() + ambiguous_dna_letters.toLowerCase()
+    );
+  }
+  __name(getAcceptedChars, "getAcceptedChars");
+  function getReplaceChars({
+    isOligo,
+    isProtein,
+    isRna,
+    isMixedRnaAndDna
+  } = {}) {
+    return isProtein ? {} : isOligo ? {} : isRna ? { t: "u" } : isMixedRnaAndDna ? {} : (
+      //just plain old dna
+      {}
+    );
+  }
+  __name(getReplaceChars, "getReplaceChars");
   function tidyUpAnnotation(_annotation, {
     sequenceData = {},
     convertAnnotationsFromAAIndices,
@@ -11417,14 +11490,6 @@ var __async = (__this, __arguments, generator) => {
     }
   }
   __name(coerceLocation, "coerceLocation");
-  function filterAminoAcidSequenceString(sequenceString, options) {
-    options = options || {};
-    if (options.includeStopCodon) {
-      return sequenceString == null ? void 0 : sequenceString.replace(/[^xtgalmfwkqespvicyhrndu.*]/gi, "");
-    }
-    return sequenceString == null ? void 0 : sequenceString.replace(/[^xtgalmfwkqespvicyhrndu]/gi, "");
-  }
-  __name(filterAminoAcidSequenceString, "filterAminoAcidSequenceString");
   function getDegenerateDnaStringFromAAString(aaString) {
     return aaString.split("").map((char) => aminoAcidToDegenerateDnaMap[char.toLowerCase()] || "nnn").join("");
   }
@@ -11436,11 +11501,10 @@ var __async = (__this, __arguments, generator) => {
       removeUnwantedChars,
       additionalValidChars,
       noTranslationData,
-      charOverrides,
       doNotProvideIdsForAnnotations,
-      proteinFilterOptions,
       noCdsTranslations,
-      convertAnnotationsFromAAIndices
+      convertAnnotationsFromAAIndices,
+      topLevelSeqData
     } = options;
     let seqData = lodashExports.cloneDeep(pSeqData);
     const response = {
@@ -11470,16 +11534,15 @@ var __async = (__this, __arguments, generator) => {
     }
     if (removeUnwantedChars) {
       if (seqData.isProtein) {
-        seqData.proteinSequence = filterAminoAcidSequenceString(
-          seqData.proteinSequence,
-          __spreadValues({ includeStopCodon: true }, proteinFilterOptions)
-        );
+        const [newSeq] = filterSequenceString(seqData.proteinSequence, __spreadValues({
+          includeStopCodon: true
+        }, topLevelSeqData || seqData));
+        seqData.proteinSequence = newSeq;
       } else {
-        seqData.sequence = filterSequenceString(
-          seqData.sequence,
-          `${additionalValidChars || ""}${seqData.isRna || seqData.isMixedRnaAndDna ? "u" : ""}`,
-          charOverrides
-        );
+        const [newSeq] = filterSequenceString(seqData.sequence, __spreadValues({
+          additionalValidChars
+        }, topLevelSeqData || seqData));
+        seqData.sequence = newSeq;
       }
     }
     if (seqData.isProtein) {
@@ -19314,7 +19377,8 @@ var __async = (__this, __arguments, generator) => {
       inclusive1BasedEnd,
       additionalValidChars,
       allowOverflowAnnotations,
-      coerceFeatureTypes
+      coerceFeatureTypes,
+      includeStopCodon
     } = options;
     [
       "isDNA",
@@ -19364,7 +19428,6 @@ var __async = (__this, __arguments, generator) => {
       response.messages.push("No sequence detected");
       sequence.sequence = "";
     }
-    let validChars;
     if (sequence.isProtein === void 0 && guessIfProtein) {
       sequence.isProtein = !guessIfSequenceIsDnaAndNotProtein(
         sequence.sequence,
@@ -19372,12 +19435,15 @@ var __async = (__this, __arguments, generator) => {
       );
     }
     if (sequence.isProtein) {
-      validChars = filterAminoAcidSequenceString(sequence.sequence);
+      const [validChars, warnings] = filterSequenceString(sequence.sequence, {
+        name: sequence.name,
+        isProtein: true,
+        additionalValidChars,
+        includeStopCodon
+      });
       if (validChars !== sequence.sequence) {
         sequence.sequence = validChars;
-        response.messages.push(
-          "Import Error: Illegal character(s) detected and removed from amino acid sequence. Allowed characters are: xtgalmfwkqespvicyhrndu"
-        );
+        response.messages.push(...warnings);
       }
       sequence.type = "PROTEIN";
       sequence.isProtein = true;
@@ -19399,12 +19465,12 @@ var __async = (__this, __arguments, generator) => {
       } else {
         sequence.type = "DNA";
       }
-      validChars = filterSequenceString(sequence.sequence, additionalValidChars);
+      const [validChars, warnings] = filterSequenceString(sequence.sequence, __spreadValues({
+        additionalValidChars
+      }, sequence));
       if (validChars !== sequence.sequence) {
         sequence.sequence = validChars;
-        response.messages.push(
-          "Import Error: Illegal character(s) detected and removed from sequence. Allowed characters are: atgcyrswkmbvdhn"
-        );
+        response.messages.push(...warnings);
       }
     }
     if (!sequence.size) {

package/package.json

@@ -1,8 +1,8 @@
 {
   "name": "@teselagen/bio-parsers",
-  "version": "0.3.9",
+  "version": "0.4.1",
   "dependencies": {
-    "@teselagen/sequence-utils": "0.3.7",
+    "@teselagen/sequence-utils": "0.3.9",
     "@teselagen/range-utils": "0.3.7",
     "@gmod/gff": "^1.2.1",
     "buffer": "^6.0.3",

package/src/utils/validateSequence.js

@@ -1,7 +1,6 @@
 import areNonNegativeIntegers from "validate.io-nonnegative-integer-array";
 import { getFeatureTypes } from "@teselagen/sequence-utils";
 import {
-  filterAminoAcidSequenceString,
   filterSequenceString,
   guessIfSequenceIsDnaAndNotProtein
 } from "@teselagen/sequence-utils";
@@ -30,7 +29,8 @@ export default function validateSequence(sequence, options = {}) {
     inclusive1BasedEnd,
     additionalValidChars,
     allowOverflowAnnotations,
-    coerceFeatureTypes
+    coerceFeatureTypes,
+    includeStopCodon
   } = options;
   [
     "isDNA",
@@ -84,7 +84,7 @@ export default function validateSequence(sequence, options = {}) {
     response.messages.push("No sequence detected");
     sequence.sequence = "";
   }
-  let validChars;
+
   if (sequence.isProtein === undefined && guessIfProtein) {
     sequence.isProtein = !guessIfSequenceIsDnaAndNotProtein(
       sequence.sequence,
@@ -93,12 +93,15 @@ export default function validateSequence(sequence, options = {}) {
   }
   if (sequence.isProtein) {
     //tnr: add code to strip invalid protein data..
-    validChars = filterAminoAcidSequenceString(sequence.sequence);
+    const [validChars, warnings] = filterSequenceString(sequence.sequence, {
+      name: sequence.name,
+      isProtein: true,
+      additionalValidChars,
+      includeStopCodon
+    });
     if (validChars !== sequence.sequence) {
       sequence.sequence = validChars;
-      response.messages.push(
-        "Import Error: Illegal character(s) detected and removed from amino acid sequence. Allowed characters are: xtgalmfwkqespvicyhrndu"
-      );
+      response.messages.push(...warnings);
     }
     sequence.type = "PROTEIN";
     sequence.isProtein = true;
@@ -126,12 +129,13 @@ export default function validateSequence(sequence, options = {}) {
       sequence.type = "DNA";
     }
 
-    validChars = filterSequenceString(sequence.sequence, additionalValidChars);
+    const [validChars, warnings] = filterSequenceString(sequence.sequence, {
+      additionalValidChars,
+      ...sequence
+    });
     if (validChars !== sequence.sequence) {
       sequence.sequence = validChars;
-      response.messages.push(
-        "Import Error: Illegal character(s) detected and removed from sequence. Allowed characters are: atgcyrswkmbvdhn"
-      );
+      response.messages.push(...warnings);
     }
   }