A cell therapy product is not a molecule in a vial. It is a population of living cells that was, in many cases, inside the patient only two weeks earlier, and it will go on changing after it is infused. This chapter explains what that means for everyone who touches the product: the scientist who grows it, the QA reviewer who releases it, the nurse who hangs the bag and the engineer who writes the software that tracks it. It then maps the WIA-BIO-005 specification onto that journey.
Most medicines fall into two familiar groups. Small-molecule drugs are chemically defined: every tablet in a lot holds the same compound, and purity can be proven by analytical chemistry. Biologics such as monoclonal antibodies are made by living cells, but the product itself is a protein that is purified away from those cells, characterised and then stored for months or years. A cell therapy product goes one step further: the cells are the active ingredient. They are harvested, sometimes genetically modified, grown, formulated and infused while still alive, and their therapeutic effect depends on what they do inside the patient afterwards: dividing, homing to tissue, killing target cells or secreting signalling molecules.
This is why the field calls them "living drugs," and it has practical consequences that run through every chapter of this book.
WIA-BIO-005 §1.1 states that the standard covers both autologous and allogeneic cell products. In an autologous product the starting material is collected from the patient who will receive it. In an allogeneic product the cells come from a different person, a donor, and a single manufacturing run can in principle supply several patients.
| Aspect | Autologous | Allogeneic |
|---|---|---|
| Starting material | The patient's own cells (for T-cell products, usually a leukapheresis collection) | Cells from a screened donor, cord blood or a cell line |
| Batch-to-patient relation | One batch, one patient | One batch may serve several patients |
| Main identity risk | Returning the wrong patient's product | Donor eligibility and immune compatibility |
| Donor screening (spec §2.4.1) | Tested as required for the collection | Full panel, including CMV antibody |
| Scheduling | Manufacturing starts after the patient is identified; the patient waits | Product can be made in advance and held in inventory ("off-the-shelf" potential) |
| Starting-material quality | Depends on the patient's disease and prior treatment | Selected donors, more consistent material |
The SDK's CellSource type also contains a third value, xenogeneic (cells of animal origin). The specification text does not describe xenogeneic products anywhere, and none of the dose ranges, release criteria or donor tests in the standard were written with them in mind. An implementation that accepts xenogeneic as input should treat it as outside the scope of WIA-BIO-005 v1.0 and route such records for manual review rather than apply human-cell rules silently.
In conventional pharmaceutical manufacturing a lot is large, and a single failed lot is a commercial loss spread across many future patients. In autologous cell therapy a lot is exactly one patient's treatment. Three things follow from this. First, there is no averaging: a batch that fails release cannot be replaced from stock, and the patient, often seriously ill, may need a new collection and another manufacturing cycle. Second, every record must carry an unbreakable link between the patient and the product, known as the chain of identity, from the moment of collection to the moment of infusion. Third, quality control sample sizes are constrained, because every millilitre used for testing is a millilitre that is not given to the patient.
The software implications are immediate. A system built for WIA-BIO-005 cannot key its records on a lot number alone; it must bind the product identifier (for example CT-2024-001, the format used in spec §8.3.1) to a pseudonymous patient identifier, and it must check that binding again at the bedside. Chapter 4 covers the data model and Chapter 6 the bedside check.
The terminology section of the specification (§1.4) defines only a handful of terms. Two of them carry the whole quality concept of the standard:
They measure different things, and a product needs both. Viability tells you whether the cells survived processing and storage; a dye-exclusion test can measure it within minutes. Potency tells you whether the surviving cells can still do their job, for example kill a tumour cell that carries the target antigen. A product can have excellent viability and poor potency (cells that are alive but exhausted or not properly engineered), and dead cells are not merely inert: they release debris and intracellular contents into the infusion. For this reason the release criteria in spec §4.2 set a floor for both, and Chapter 3 explains how each is measured and why different methods can give different answers for the same sample.
Section 2 of the specification describes five product families. They differ in where the cells come from, how long it takes to make them, what dose unit is used and which markers prove that the product is what its label says. The table below summarises what the specification itself states. Where the specification is silent, the table says so rather than filling the gap.
| Family | Source (spec) | Typical timeline (spec) | Dose as stated in spec §2 | Primary identity markers |
|---|---|---|---|---|
| CAR-T | Autologous T cells, collected by leukapheresis | 10–14 days manufacturing | 1×10⁶ – 5×10⁶ cells/kg | CD3+, CAR+ (release panel §4.1.1) |
| TIL | Tumour biopsy | 4–6 weeks expansion | 1×10¹⁰ – 1×10¹¹ total cells | Not specified separately; T-cell panel (CD3+) is the natural fit |
| CAR-NK | Peripheral blood, cord blood or iPSCs | Not stated | 1×10⁷ – 1×10⁹ cells/kg (see note below) | CD56+, CD3− (NK panel §4.1.1) |
| MSC | Bone marrow, adipose tissue, umbilical cord | Passages P2–P5 for clinical use | 1×10⁶ – 2×10⁶ cells/kg | CD73+, CD90+, CD105+; CD34−, CD45−, CD11b−, CD19−, HLA-DR− |
| iPSC | Reprogrammed somatic cells | 21–28 days for reprogramming | Not stated | SSEA-4, TRA-1-60, OCT4, NANOG (pluripotency) |
Reading the dose column correctly. These figures describe what WIA-BIO-005 encodes. They are not treatment instructions; the approved product label and the treating centre's protocol always take precedence. Two entries need particular care. The TIL figure is a total dose per patient, not a per-kilogram dose, and must not be divided by weight or compared against per-kilogram ranges. The CAR-NK range is high compared with the doses reported in published early-phase CAR-NK studies, which have generally been lower by one to several orders of magnitude; treat the specification's range as an encoded bound, not as a typical dose. Chapter 5 shows where the reference SDK handles both of these incorrectly.
CAR-T products are the best-known family and the one for which the specification gives the most detail. The standard lists six target antigens: CD19, CD22, BCMA, CD30, HER2 and GD2. The first four are expressed on blood cancers (B-cell leukaemias and lymphomas for CD19 and CD22, multiple myeloma for BCMA, Hodgkin and some T-cell lymphomas for CD30). HER2 and GD2 are solid-tumour antigens, HER2 in some carcinomas and GD2 in neuroblastoma and other tumours of neural-crest origin. The list mixes established targets with investigational ones, and it should be read as the set of antigens the standard's data model is designed to record, not as a list of approved indications.
Tumour-infiltrating lymphocytes are T cells that have already found their way into a tumour. Instead of engineering a new receptor, TIL manufacturing isolates these cells from a surgical or biopsy specimen and expands them to very large numbers. The long expansion period (4–6 weeks in the specification) and the total-dose convention are the two features that distinguish TIL from CAR-T in a tracking system.
Natural killer cells can carry the same kind of engineered receptor as T cells. The specification lists three sources (peripheral blood, cord blood and iPSCs) and two advantages, "off-the-shelf potential" and "lower toxicity." Both are stated as potential, and this book treats them the same way (see Section 1.3).
Mesenchymal stem (stromal) cells and induced pluripotent stem cells are covered in Section 1.4. They are included in the standard because they share the same manufacturing, cryopreservation and release logic, even though their mechanism of action has nothing to do with an engineered receptor.
The reference TypeScript SDK defines CellType with nine values: the five families above plus HSC (haematopoietic stem cells), DC (dendritic cells), Treg (regulatory T cells) and Other. The dose request interface in spec §8.2.1 allows only the original five. Nothing in the specification defines dose ranges, identity panels or release criteria for the four additional types.
This is a classic interface drift, and it has a concrete effect. When calculateDose in the SDK receives a cell type it has no specific branch for (iPSC, HSC, DC, Treg or Other), it falls back silently to a default window of 1×10⁶ – 1×10⁷ cells/kg and reports the result as if it were meaningful. An implementer has three honest options:
What an implementer should not do is show a green "within range" indicator based on the fallback window. Chapter 5 returns to this when it discusses wrapping the SDK rather than forking it.
A chimeric antigen receptor is a synthetic protein that joins, in a single molecule, a part that recognises a target on the outside of a cell and a part that switches on the T cell from the inside. "Chimeric" refers to this joining of domains from different natural proteins. A typical CAR has four segments:
| Segment | Usual origin | Function |
|---|---|---|
| Antigen-recognition domain | Single-chain variable fragment (scFv) derived from an antibody | Binds the target antigen, for example CD19, on the tumour cell surface |
| Hinge (spacer) | Segments of CD8α, CD28 or immunoglobulin | Gives the binding domain reach and flexibility |
| Transmembrane domain | Commonly CD8α or CD28 | Anchors the receptor in the cell membrane |
| Intracellular signalling domains | CD3ζ plus one co-stimulatory domain (CD28 or 4-1BB) in second-generation designs | Delivers the activation signal (CD3ζ) and the survival and proliferation signal (co-stimulation) |
Two properties of this design explain why CAR-T works and why it is hard to control. First, recognition is independent of MHC. A natural T-cell receptor sees fragments of proteins presented on HLA molecules, and tumours often escape by reducing HLA expression. A CAR binds the intact surface antigen directly, as an antibody would, and then turns that binding into a full T-cell response. Second, the co-stimulatory domain lets the engineered cell multiply after it meets its target. The infused dose is therefore a starting number, not a fixed exposure. This is why a weight-based dose of a few million cells per kilogram can be effective, and why the peak of expansion, typically in the first one to two weeks, is also the peak risk window for cytokine release syndrome.
CD19 is an almost ideal first target. It is present on the great majority of B-cell leukaemias and lymphomas, it is expressed throughout B-cell development, and it is absent from haematopoietic stem cells and from non-haematopoietic tissues. Killing every CD19-positive cell therefore removes the cancer together with the patient's normal B cells, but spares the stem cells that will eventually rebuild the B-cell compartment. The resulting "B-cell aplasia" and low antibody levels are an expected on-target effect, which is managed with infection monitoring and, when needed, immunoglobulin replacement (Chapter 7). For implementers this is a useful example of a finding that is expected and must not be raised as an unexpected adverse event by default.
Using another person's T cells is difficult, because their natural T-cell receptors can recognise the recipient's tissues as foreign and cause graft-versus-host disease. NK cells do not use a rearranged T-cell receptor and are not a major driver of graft-versus-host disease, so a product made from a donor, from cord blood or from an iPSC line could be manufactured in advance, tested, frozen and supplied to many patients. That is what the specification means by "off-the-shelf potential" in §2.2.1. The word potential matters: persistence of NK cells in the recipient, immune rejection of donor cells and consistency between batches remain active research questions, and the specification does not claim otherwise. The phrase "lower toxicity" should likewise be read as a qualitative expectation, not as an established comparative result.
MSCs are adherent, fibroblast-like cells that can be isolated from bone marrow, adipose tissue and umbilical cord, the three sources listed in spec §2.3.1. Unlike CAR-T, they are not engineered to kill anything. Their proposed effects are largely paracrine and immunomodulatory, which is also why potency assays for MSC products are harder to define than a cytotoxicity test. The International Society for Cell & Gene Therapy (ISCT) has recommended that such cells be called "mesenchymal stromal cells" unless stem-cell properties have actually been demonstrated; the specification uses "mesenchymal stem cells," and this book uses the abbreviation MSC for both.
The identity definition in the specification follows the minimal criteria published by ISCT in 2006, which remain the common reference point for MSC characterisation:
The specification also restricts clinical material to passages P2–P5. A passage is one cycle of detaching adherent cells and reseeding them into fresh culture vessels. Each passage increases cell number, but cells that are passaged many times drift in their properties and approach senescence, so the passage window is a simple and auditable proxy for "not too early to have enough cells, not too late to have changed." MSC products also carry a stricter viability requirement at the product-family level, above 90% in §2.3.1, than the general release floor of 85% in §4.2. A data model has to allow a product-specific criterion to override the general one.
iPSCs are made by reprogramming ordinary adult cells, such as skin fibroblasts or blood cells, back into a pluripotent state, in which they can in principle give rise to any cell type of the body. The discovery that four transcription factors are enough to do this, first shown in mouse cells in 2006 by Takahashi and Yamanaka, is why the specification refers to "Yamanaka factors": Oct4, Sox2, Klf4 and c-Myc. The specification gives a reprogramming timeline of 21–28 days and four pluripotency markers for characterisation: SSEA-4, TRA-1-60, OCT4 and NANOG.
One point is essential for anyone modelling iPSC products: undifferentiated iPSCs are not the therapeutic product. Pluripotent cells can form teratomas, so a clinical product is a differentiated derivative (for example NK cells, retinal pigment epithelium or dopaminergic neurons) made from a characterised iPSC line. The pluripotency markers listed in the specification characterise the master cell line; in the final product, residual cells that still express them are an impurity to be minimised, not an identity to be confirmed. WIA-BIO-005 v1.0 does not yet state release criteria for iPSC-derived products, and the SDK's fallback dose window for iPSC (Section 1.2) has no basis in the specification. A system that tracks iPSC-derived products should record both the parent line and the differentiated product as separate, linked entities.
The term vein-to-vein describes the complete path of an autologous product: from the patient's vein at collection back to the patient's vein at infusion. Each step belongs to a different department, and often to a different organisation: an apheresis unit, a manufacturing site that may be on another continent, a courier, a hospital cell-processing laboratory, a pharmacy and a ward. Figure 1-1 shows the chain with the specification section that governs each step.
Reading the chain step by step:
Two threads run through all eight steps. The chain of identity is the unbroken association between patient, collection, batch, bags and infusion record: the answer to "whose cells are these?" at any moment. The chain of custody is the record of who held the material, where, under what conditions and for how long: the answer to "has anything happened to these cells that we do not know about?" Temperature excursions during shipping, an unlogged hand-over or a label reprinted without reconciliation are chain-of-custody failures even when the right product reaches the right patient. Most of the data model in Chapter 4 exists to make these two chains machine-checkable.
WIA-BIO-005 is published as version 1.0.0, dated 2025-12-26, with status "Active," under the MIT licence. Its stated purpose (§1.1) is to define standards for the development, manufacturing, quality assurance and clinical use of cellular therapies. Section 1.2 lists six areas in scope:
It is worth being precise about what kind of document this is. WIA-BIO-005 is a WIA industry standard. It collects process parameters, acceptance criteria and data structures in one place and defines what a conforming software system must contain. It is not a regulation, and it is not recognised by the FDA, the EMA or any other authority as a substitute for their requirements. Where the standard quotes clinical values (doses, chemotherapy regimens, drug doses for toxicity management), those values are what the standard encodes. They do not replace the approved product label or the protocol of the treating institution, and this book repeats that caution wherever such values appear.
The single most important sentence for implementers is in §8.1: any WIA-BIO-005 compliant system must include five components.
| Component (§8.1) | Stated role | Where it acts on the chain | Covered in |
|---|---|---|---|
| Dose Calculator | Calculate patient-specific doses | Release and infusion | Chapter 5; workflow in Chapter 6 |
| Quality Assessor | Evaluate release criteria | In-process checkpoints and lot release | Chapter 3 (tests); Chapter 5 (algorithm) |
| Manufacturing Tracker | Monitor batch progress | Collection to cryopreservation | Chapter 2 (process); Chapter 5 (code); Chapter 6 (state machine) |
| Safety Monitor | Track adverse events | Infusion and follow-up | Chapter 7 |
| Compliance Checker | Verify regulatory requirements | Every step | Chapter 8 |
The specification defines the API shape of the first three in §8.2 and a canonical JSON product record in §8.3. It lists error codes C001–C006 in §8.4; the reference SDK adds four more. It does not define an API for the Safety Monitor or the Compliance Checker, so those are described in this book on the basis of the SDK and of the regulatory requirements in §9.
The standard is distributed with a TypeScript SDK (@wia/bio-005, class CellularTherapySDK), a command-line tool (wia-bio-005) and an interactive simulator. They are useful, and later chapters use them extensively, but they are not the specification. They sometimes disagree with it and with each other: the extra cell types discussed in Section 1.2 are one example, and the quality grading rules and error code sets discussed in Chapter 5 are others. This book follows one rule throughout: the specification text is the reference; where it is itself inconsistent or out of date, the book says so explicitly and gives the corrected reading with its source. Where the SDK deviates, the book shows how to wrap it so that the integration behaves according to the corrected reading. The repository also contains four generic "PHASE" documents that do not contain cell-therapy requirements; this book does not treat them as normative, and Chapter 8 uses only their general ideas about conformance evidence.
Section 10.3 names four related WIA standards, which together describe where WIA-BIO-005 sits in a larger clinical system:
WIA-BIO-005 itself focuses on the product: how it is made, tested, stored, dosed and followed up. Chapter 7 shows how its components exchange events with systems built on these neighbouring standards.
| Spec section | Content | Main chapter in this book |
|---|---|---|
| §1 Introduction | Purpose, scope, terminology | Chapter 1 |
| §2 Cell types and sources | Product families, donor screening, starting material | Chapters 1 and 2 |
| §3 Manufacturing processes | Facility, equipment, CAR-T process, in-process controls, CPPs | Chapter 2 |
| §4 Quality control standards | Identity, viability, potency, sterility, endotoxin, mycoplasma, stability | Chapter 3 |
| §5 Cryopreservation protocols | Formulation, controlled-rate freezing, thaw, storage | Chapter 6 |
| §6 Clinical administration | Dose calculation, lymphodepletion, infusion, post-infusion monitoring | Chapters 5 and 6 |
| §7 Safety monitoring | CRS, ICANS, other adverse events | Chapter 7 |
| §8 Implementation guidelines | Required components, API, data format, error codes | Chapters 4 and 5 |
| §9 Regulatory compliance; §10 References | FDA, EMA and ICH mapping; quality standards | Chapter 8 |
The chapters follow the vein-to-vein chain and then turn to the software that has to follow it:
The book is written for two groups of readers who need to understand each other.
If you are an engineer (building a LIMS or MES integration, a hospital workflow or a conformance test suite), read this chapter fully, then Chapters 4 and 5 for the data and code, then Chapter 6 for the workflow. Go back to Chapters 2 and 3 whenever a field or a threshold in the code seems arbitrary; the reason for it is almost always there. Chapter 8 tells you what evidence you will need to show that your system conforms.
If you work in manufacturing, QC, QA or the clinic, read Chapters 1 to 3 in order, then Chapters 6 and 7. Chapters 4 and 5 are still worth skimming: they show how your batch records, certificates and assessments will be represented, and where the software's defaults may differ from your procedures. The known-gaps section at the end of Chapter 5 is written for you as much as for developers.
Chapter summary. A cell therapy product is a population of living cells whose quality is defined above all by viability and potency, and an autologous product is a batch of one that must stay bound to its patient from collection to infusion. WIA-BIO-005 describes five product families (CAR-T, TIL, CAR-NK, MSC and iPSC) along the vein-to-vein chain, and requires every conforming system to provide five components: Dose Calculator, Quality Assessor, Manufacturing Tracker, Safety Monitor and Compliance Checker. The reference SDK extends the specification in places without defining the rules behind the extensions; an implementer must notice these gaps and close them deliberately rather than inherit silent defaults.