Getting Started with VigyanLLM

Design lab-ready primer and probe pairs with our 24-step biophysical validation pipeline. From sequence input to an audit-ready PDF report in minutes.

How the Pipeline Works

VigyanLLM's pipeline is a fully automated 24-step primer and probe design engine. It takes a DNA sequence or gene name as input and produces lab-ready primer pairs with an audit-grade PDF report. The entire pipeline runs server-side — the browser submits the job and polls for results.

Architecture Overview

When you submit a target sequence, the job is dispatched to a Flask backend running Primer3, NCBI BLAST, and the SantaLucia nearest-neighbour thermodynamics engine. Each of the 24-steps is executed sequentially. If any step fails (e.g., BLAST is unreachable), the pipeline marks that job as failed and returns a partial result. Results are stored server-side and fetched by the frontend via polling.

Pipeline at a Glance

Phase A

5

Target Selection & Filtering

Phase B

4

Primer Design & Thermodynamics

Phase C

3

Specificity & Alignment

Phase D

7

Secondary Structure & Validation

Phase E

3

Ranking & Report

Step-by-Step Pipeline Flow

Step Phase Description
1AFetch reference transcript & genomic sequence from NCBI
2AFilter low-complexity & repeat-masked regions
3AIdentify conserved & unique regions for primer placement
4AExclude known SNP-dense regions from target window
5ASelect optimal amplification window based on GC profile
6BRun Primer3 with configured Tm, length, and GC% parameters
7BApply 3′ GC clamp & end-stability filters
8BCalculate Tm via SantaLucia 1998 nearest-neighbour model
9BEnforce ΔTm ≤ 1.5°C between forward & reverse
10CBLAST each candidate against reference genome (GRCh38/hg38)
11CFilter off-targets with ≤ 2 mismatches in last 5 bp at 3′ end
12CCross-reference amplicon against known pseudogenes
13DPredict hairpin secondary structure (ΔG > −2.0 kcal/mol)
14DPredict self-dimer formation (ΔG > −8.0 kcal/mol)
15DPredict cross-dimer between Fwd & Rev (ΔG > −8.0 kcal/mol)
16DMask primers overlapping with Dfam repeat elements
17DCross-reference positions against dbSNP common variants
18DEvaluate multiplex compatibility (cross-pair ΔG > −5.0)
19DDesign TaqMan probe (if enabled): Tm 68–70°C, no G at 5′
20ECompute composite score (0–100) per primer pair
21ERank pairs & filter to top 10 best candidates
22EGenerate PDF audit report with full traceability

How Scoring Works

Each of the 24-steps contributes a weighted sub-score to the final 0–100 rating. Thermodynamic penalties (hairpin, self-dimer, cross-dimer) carry the heaviest weight because they directly affect PCR efficiency. BLAST off-target count and SNP overlap also reduce the score. Pairs scoring 80+ are marked green (lab-ready), 60–79 are yellow (requires verification), 40–59 are orange (use with caution), and <40 are red (discard).

User Workflow

1Enter Your Sequence

Paste a DNA template sequence or a gene name (e.g., TP53, BRCA1). The pipeline auto-fetches the transcript from NCBI. Supported length: 50–10,000 bp.

2Configure Parameters

Set primer Tm (default 62°C), amplicon size range, and design mode (Standard PCR, Bisulfite MSP, or Multiplex). Enable TaqMan probe design or NGS adapter tailing as needed.

Optional

3Run Pipeline

Click Run Pipeline. The 24-step engine executes server-side in sequence. The frontend polls for progress and updates the UI as each phase completes. Typical runtime: 30–60 seconds depending on target length and server load.

~30–60 seconds

4Review Results

Each primer pair is scored 0–100. Green (80+) pairs are lab-ready. Yellow/orange pairs need verification. Red pairs should be discarded. Expand any row to see full metrics: Tm, GC%, hairpin ΔG, self-dimer ΔG, cross-dimer ΔG, and BLAST off-target hits.

5Export & Report

Download results as CSV, JSON, or a comprehensive PDF audit report. Order-ready reports for IDT and Twist Bioscience are also available for direct synthesis ordering.

PDF includes full audit trail

Try the Pipeline Now

No account required for the demo. Enter a sequence and see the 24-step engine in action.