agentic-flow 1.9.0 → 1.9.1

package/agentic-flow/examples/crispr-cas13-pipeline/docs/flowcharts/offtarget-flowchart.md

@@ -1,317 +0,0 @@
-# Off-Target Prediction Module - Flowchart
-
-## Main Off-Target Prediction Workflow
-
-```mermaid
-flowchart TD
-    Start([Start: gRNA Input]) --> ValidateRNA[Validate gRNA<br/>Length & Bases]
-
-    ValidateRNA --> LoadResources[Load Transcriptome<br/>& ML Model]
-
-    LoadResources --> BuildIndex{K-mer Index<br/>Built?}
-
-    BuildIndex -->|No| CreateIndex[Build K-mer Index<br/>from Transcriptome]
-    BuildIndex -->|Yes| ExtractSeed
-
-    CreateIndex --> ExtractSeed[Extract Seed Region<br/>Last 12 bases]
-
-    ExtractSeed --> FindSeeds[Find Seed Matches<br/>in K-mer Index]
-
-    FindSeeds --> CheckMatches{Seed Matches<br/>Found?}
-
-    CheckMatches -->|No| NoTargets[Return Empty<br/>Result Set]
-    CheckMatches -->|Yes| VerifyFull[Verify Full gRNA<br/>Alignment]
-
-    VerifyFull --> FilterMM[Filter by<br/>Max Mismatches]
-
-    FilterMM --> CheckCandidates{Candidates > 0?}
-
-    CheckCandidates -->|No| NoTargets
-    CheckCandidates -->|Yes| ParallelFeatures{Parallel Feature<br/>Extraction}
-
-    ParallelFeatures --> Batch1[Batch 1:<br/>Extract Features]
-    ParallelFeatures --> Batch2[Batch 2:<br/>Extract Features]
-    ParallelFeatures --> Batch3[Batch 3:<br/>Extract Features]
-
-    Batch1 --> MergeFeatures[Merge Feature<br/>Vectors]
-    Batch2 --> MergeFeatures
-    Batch3 --> MergeFeatures
-
-    MergeFeatures --> Predict[Gradient Boosting<br/>Prediction]
-
-    Predict --> ScoreSites[Calculate Off-Target<br/>Scores & Probabilities]
-
-    ScoreSites --> FilterThreshold[Filter by Score<br/>& Expression]
-
-    FilterThreshold --> RankSites[Rank by Composite<br/>Score]
-
-    RankSites --> Annotate[Add Gene Annotations<br/>& Pathways]
-
-    Annotate --> GenStats[Generate Summary<br/>Statistics]
-
-    GenStats --> End([End: Ranked<br/>Off-Targets])
-
-    NoTargets --> End
-
-    style Start fill:#e1f5e1
-    style End fill:#ffe1e1
-    style ParallelFeatures fill:#fff4e1
-    style CheckMatches fill:#fff4e1
-    style CheckCandidates fill:#fff4e1
-    style Predict fill:#e1f0ff
-```
-
-## Feature Extraction Detail
-
-```mermaid
-flowchart TD
-    StartFeat([Start: Candidate Site]) --> SeqFeatures[Sequence Features]
-
-    SeqFeatures --> CalcMM[Calculate Mismatch<br/>Distribution]
-    CalcMM --> CalcGC[Calculate GC<br/>Content]
-    CalcGC --> CalcHomopoly[Calculate Homopolymer<br/>Runs]
-
-    CalcHomopoly --> StructFeatures[Structure Features]
-
-    StructFeatures --> PredictStruct[Predict RNA<br/>Secondary Structure]
-    PredictStruct --> CalcEnergy[Calculate Free<br/>Energy]
-    CalcEnergy --> CalcAccess[Calculate<br/>Accessibility]
-
-    CalcAccess --> ExprFeatures[Expression Features]
-
-    ExprFeatures --> GetTPM[Get Transcript<br/>TPM Value]
-    GetTPM --> GetCoverage[Get Local Read<br/>Coverage]
-
-    GetCoverage --> ChromFeatures{Chromatin Data<br/>Available?}
-
-    ChromFeatures -->|Yes| GetChromatin[Get Chromatin<br/>Accessibility]
-    ChromFeatures -->|No| DefaultChrom[Use Default<br/>Value]
-
-    GetChromatin --> EvoFeatures
-    DefaultChrom --> EvoFeatures[Evolutionary Features]
-
-    EvoFeatures --> GetConservation[Get PhyloP<br/>Conservation Score]
-
-    GetConservation --> DistFeatures[Distance Features]
-
-    DistFeatures --> CalcJunction[Calculate Distance to<br/>Exon Junction]
-
-    CalcJunction --> ThermoFeatures[Thermodynamic Features]
-
-    ThermoFeatures --> CalcAffinity[Calculate Binding<br/>Affinity]
-
-    CalcAffinity --> PackageVector[Package Feature<br/>Vector 15-20 dims]
-
-    PackageVector --> EndFeat([Return: Feature<br/>Vector])
-
-    style StartFeat fill:#e1f5e1
-    style EndFeat fill:#ffe1e1
-    style PredictStruct fill:#e1f0ff
-    style ChromFeatures fill:#fff4e1
-```
-
-## Gradient Boosting Training
-
-```mermaid
-flowchart TD
-    StartTrain([Start: Training Data]) --> InitModel[Initialize Model<br/>Base Score = Mean]
-
-    InitModel --> InitPred[Initialize Predictions<br/>with Base Score]
-
-    InitPred --> LoopTrees{Tree Iteration<br/>< NumTrees?}
-
-    LoopTrees -->|Yes| CalcResidual[Calculate Residuals<br/>Label - Prediction]
-
-    CalcResidual --> Subsample[Subsample Training<br/>Data 80%]
-
-    Subsample --> BuildTree[Build Regression Tree<br/>Max Depth = 6]
-
-    BuildTree --> TreeLoop{Split Node?}
-
-    TreeLoop -->|Yes| FindBestSplit[Find Best Split<br/>Max Variance Reduction]
-    FindBestSplit --> SplitNode[Split into Left/Right<br/>Children]
-    SplitNode --> TreeLoop
-
-    TreeLoop -->|No| CreateLeaf[Create Leaf Node<br/>Prediction = Mean]
-
-    CreateLeaf --> AddTree[Add Tree to Model]
-
-    AddTree --> UpdatePred[Update Predictions<br/>+= LR × TreePred]
-
-    UpdatePred --> CalcLoss[Calculate Training<br/>Loss MSE]
-
-    CalcLoss --> LogProgress{Iteration % 10<br/>== 0?}
-
-    LogProgress -->|Yes| LogInfo[Log Training<br/>Progress]
-    LogProgress -->|No| NextTree
-
-    LogInfo --> NextTree[Tree++]
-    NextTree --> LoopTrees
-
-    LoopTrees -->|No| CalcImportance[Calculate Feature<br/>Importances]
-
-    CalcImportance --> EndTrain([Return: Trained<br/>Model])
-
-    style StartTrain fill:#e1f5e1
-    style EndTrain fill:#ffe1e1
-    style BuildTree fill:#e1f0ff
-    style LoopTrees fill:#fff4e1
-    style TreeLoop fill:#fff4e1
-    style LogProgress fill:#fff4e1
-```
-
-## Off-Target Scoring & Ranking
-
-```mermaid
-flowchart TD
-    StartScore([Start: Candidates]) --> LoopCandidates{For Each<br/>Candidate}
-
-    LoopCandidates --> ExtractFeat[Extract Feature<br/>Vector]
-
-    ExtractFeat --> GBPredict[Gradient Boosting<br/>Prediction]
-
-    GBPredict --> Sigmoid[Apply Sigmoid<br/>to Get Probability]
-
-    Sigmoid --> CalcCI[Calculate Confidence<br/>Interval]
-
-    CalcCI --> StoreResult[Store Scored<br/>Result]
-
-    StoreResult --> NextCandidate[Next Candidate]
-    NextCandidate --> LoopCandidates
-
-    LoopCandidates -->|Done| FilterScore{Filter by<br/>Score Threshold}
-
-    FilterScore --> FilterExpr{Filter by<br/>Expression}
-
-    FilterExpr --> CalcComposite[Calculate Composite<br/>Score with Expression]
-
-    CalcComposite --> AddAnnotations[Add Gene Annotations<br/>Biotype, Pathways]
-
-    AddAnnotations --> SortDesc[Sort by Composite<br/>Score Descending]
-
-    SortDesc --> ApplyTopK{Limit to Top K?}
-
-    ApplyTopK -->|Yes| TakeTopK[Take Top K<br/>Results]
-    ApplyTopK -->|No| AllResults
-
-    TakeTopK --> EndScore
-    AllResults[Use All Results] --> EndScore([Return: Ranked<br/>Off-Targets])
-
-    style StartScore fill:#e1f5e1
-    style EndScore fill:#ffe1e1
-    style GBPredict fill:#e1f0ff
-    style FilterScore fill:#fff4e1
-    style FilterExpr fill:#fff4e1
-    style ApplyTopK fill:#fff4e1
-```
-
-## RNA Secondary Structure Prediction (Vienna RNA)
-
-```mermaid
-flowchart TD
-    StartRNA([Start: RNA<br/>Sequence]) --> InitDP[Initialize DP Matrix<br/>n × n]
-
-    InitDP --> LoopLength{For Length<br/>2 to n}
-
-    LoopLength --> LoopI{For i = 0 to<br/>n - length}
-
-    LoopI --> CalcJ[j = i + length]
-
-    CalcJ --> CaseUnpaired[Case 1: i Unpaired<br/>dp[i][j] = dp[i+1][j]]
-
-    CaseUnpaired --> LoopK{For k = i+1<br/>to j}
-
-    LoopK --> CanPair{Can Pair<br/>seq[i], seq[k]?}
-
-    CanPair -->|Yes| GetEnergy[Get Base Pair<br/>Energy]
-    GetEnergy --> CalcPaired[Calculate Paired<br/>Energy]
-    CalcPaired --> UpdateDP[Update dp[i][j]<br/>if Better]
-
-    CanPair -->|No| NextK
-    UpdateDP --> NextK[k++]
-
-    NextK --> LoopK
-
-    LoopK -->|Done| NextI[i++]
-    NextI --> LoopI
-
-    LoopI -->|Done| NextLength[length++]
-    NextLength --> LoopLength
-
-    LoopLength -->|Done| GetMFE[MFE = dp[0][n-1]]
-
-    GetMFE --> Traceback[Traceback to Get<br/>Structure]
-
-    Traceback --> EndRNA([Return: Structure<br/>& Free Energy])
-
-    style StartRNA fill:#e1f5e1
-    style EndRNA fill:#ffe1e1
-    style LoopLength fill:#fff4e1
-    style LoopI fill:#fff4e1
-    style LoopK fill:#fff4e1
-    style CanPair fill:#fff4e1
-```
-
-## Error Handling & Resource Management
-
-```mermaid
-flowchart TD
-    Operation[Execute<br/>Off-Target Operation] --> CheckMem{Memory Usage<br/>> 90%?}
-
-    CheckMem -->|Yes| ReduceCandidates[Reduce Candidate Set<br/>Stricter Filtering]
-    CheckMem -->|No| CheckCandCount
-
-    ReduceCandidates --> CheckCandCount{Candidates<br/>> 10,000?}
-
-    CheckCandCount -->|Yes| FilterByExpr[Filter by Expression<br/>Keep High TPM Only]
-    CheckCandCount -->|No| Proceed
-
-    FilterByExpr --> Proceed[Proceed with<br/>Analysis]
-
-    Proceed --> TryOperation{Try Operation}
-
-    TryOperation -->|Success| Success([Success])
-
-    TryOperation -->|Model Load Error| CheckFile{Model File<br/>Exists?}
-    TryOperation -->|OOM Error| SwitchBatch[Reduce Batch Size]
-    TryOperation -->|Invalid Input| LogReject[Log & Reject<br/>Invalid gRNA]
-
-    CheckFile -->|No| Error1[Error: Model<br/>Not Found]
-    CheckFile -->|Yes| Error2[Error: Corrupted<br/>Model]
-
-    SwitchBatch --> RetryOp[Retry with Smaller<br/>Batches]
-    RetryOp --> Success
-
-    Error1 --> Fail([Fail])
-    Error2 --> Fail
-    LogReject --> Fail
-
-    style Success fill:#e1f5e1
-    style Fail fill:#ffe1e1
-    style CheckMem fill:#fff4e1
-    style CheckCandCount fill:#fff4e1
-    style TryOperation fill:#fff4e1
-    style CheckFile fill:#fff4e1
-```
-
----
-
-## Complexity Summary
-
-| Operation | Time Complexity | Space Complexity |
-|-----------|----------------|------------------|
-| K-mer Index Building | O(T × L) | O(T × L / K) |
-| Candidate Discovery | O(K × log T) | O(K) |
-| Feature Extraction | O(K × n²) | O(K × F) |
-| Gradient Boosting Training | O(N × F × T × D) | O(T × D) |
-| Prediction | O(K × T × D) | O(K × F) |
-| Full Pipeline | O(T × L + K × n²) | O(T × L + K × F) |
-
-Where:
-- T = transcriptome size
-- L = average transcript length
-- K = number of candidates
-- n = gRNA length
-- F = feature dimensions
-- N = training samples
-- D = tree depth