CHAPTER 2 · WHAT'S ALREADY WORKING
You already built the lab
Thermo Fisher Platform for Science captures the science. The investment is sunk · the workflow is regulated · we don't touch it.
SITUATION TODAY
What lives in PfS today (4 sites · ~40 instruments · audit-grade)
Every sample, every result, every signature is captured. The data is there. What it lacks is the enterprise-wide layer that reasons across it. PfS is — by design — a lab system, not a decisioning system.
12,400
Samples logged / year
Across MG · SG · Embelex
3,100
Experiments run / year
Captured procedurally in ELN
8,900
Test results / year
Numeric + pass/fail · spec-bounded
14,200
Instrument hours / year
~40 instruments · 4 sites
$14M
Sunk PfS investment
Capex + 3 years of licence + integration
18 mo + $9M
Recertification cost of replacing
GxP requalification · scientist retraining
BENEFIT
Keep the $14M investment · add the intelligence layer for $0 CapEx
Lab unchanged · audit trail preserved · scientists work as today · enterprise gains everything
THERMO FISHER · PLATFORM FOR SCIENCE · 5 MODULES IN SCOPE
LIMS
Core LIMS
Samples · workflows · batches
AWS · AD-tenanted · 21 CFR 11
✓
ELN
Core ELN
Procedural experiments
Native LIMS-integrated
✓
SDMS
Core SDMS
Raw instrument files
Vendor-neutral capture
✓
CDS
Chromeleon CDS
GC/LC chromatography
Method library harmonised
✓
CONNECT
Core Connect
OData v4 open API
The one door SCIKIQ uses
✓
↑ SCIKIQ reads + writes via this one door ↑
HOW IT WORKS
How SCIKIQ uses PfS (without touching it)
1
Read via OData v4
SCIKIQ polls Core Connect every 5 minutes for new/updated entities (Samples · Experiments · Results · Batches). Webhook for instant result-created events.
▣ OData v4 · OAuth 2.0 · ~12k entities/day in steady state
→
2
Project to the Knowledge Graph
Each PfS entity becomes a node; lineage edges link Sample→Experiment→Result→Batch→Formulation→Material→Supplier.
▣ Property graph · entity resolution · ontology mapping
→
3
Write enriched signals back
Three user-defined fields are added to the PfS Sample entity — Similarity_Cluster, Sustainability_Score, Customer_Fit_Score — refreshed weekly. Scientists see them in their normal LIMS view.
▣ OData PATCH · UDF write-back · no schema migration
BEFORE / AFTER
What changes for the lab (spoiler: nothing)
| Dimension | Today | With SCIKIQ + Thermo | Δ |
|---|---|---|---|
| Scientist UI / login | Core ELN / LIMS | Core ELN / LIMS (unchanged) | 0 |
| Audit trail · 21 CFR 11 | Thermo-owned | Thermo-owned (unchanged) | 0 |
| Sample / experiment / batch flow | PfS workflow | PfS workflow (unchanged) | 0 |
| Net-new PfS fields visible to scientist | — | 3 UDFs (Similarity · Sustainability · Customer-fit) | +3 helpful |
| Net-new infra / CapEx in the lab | — | $0 | 0 |
WHO WINS
"Zero risk to the system of record. Zero retraining for scientists. The lab they know stays the lab they use — SCIKIQ adds intelligence on top, not inside."
— AD R&D-IT (operator of PfS) · CIO org · Cleveland
⚠ IF WE DON'T
Why this is the only door we should use
- Replacing PfS = 18-month GxP requalification + $9M re-licence + scientist retraining
- Bypassing PfS into raw instrument files = breaks audit trail, fails ESPR substantiation
- Asking Thermo to add AI = locks AVY to one vendor's roadmap and pace
5PfS modulesLIMS · ELN · SDMS · CDS · Connect
$14Msunk investmentKept · not duplicated
$0CapExSCIKIQ sits above · not inside
12kentities / dayRead via OData v4 in steady state
Five PfS modules. Four R&D sites. ~40 instruments. This is built right. The Thermo investment is sunk and working. SCIKIQ uses the one open door — Core Connect's OData v4 API — and never asks the lab to change.
Don't break the lab. Build above it.
AI / ML / LLM IN THIS CHAPTER
DATA SOURCES