# Human vs humanized vs chimeric antibodies: the origin of every FDA-approved antibody

> Of 159 antibody products the FDA licensed by September 2026, 52% are humanized, 34% fully human, 9% chimeric and 5% murine, by ProteinIQ's label-by-label count. The fully human share has held near one-third since 2015. Free CSV.

The FDA had licensed **159 monoclonal antibody products** by September 25, 2026, and by our count 83 of them are humanized (52%), 54 fully human (34%), 14 chimeric (9%) and 8 murine (5%). We classified every product in the [FDA's Purple Book list of licensed monoclonal antibodies](/guides/how-many-monoclonal-antibodies-are-fda-approved) from the origin its prescribing information states, falling back on its international nonproprietary name (INN) where the label is silent.

The mix has changed less than the usual story suggests. Fully human antibodies went from none before 2002 to about a third of all licensed antibodies by 2015, and they have stayed near that share since. Humanized antibodies, not human ones, account for most recent approvals, largely because 16 of the 20 [bispecific antibodies](/guides/approved-bispecific-antibodies) and [antibody-drug conjugates](/guides/fda-approved-adcs) licensed since 2020 are humanized.

## What do murine, chimeric, humanized and human mean?

The four words describe how much of an antibody's sequence came from a mouse. The first therapeutic antibody, muromonab-CD3, was a mouse protein made by the hybridoma method that Köhler and Milstein published in 1975. Patients' immune systems treated it as foreign: in its early kidney-transplant trials, 86% of 43 patients made IgG antibodies against it, and those antibodies could neutralize the drug.

Each later origin removed more of the mouse. A chimeric antibody joins the mouse [variable domains](/guides/antibody-variable-regions), the part that binds the target, to human constant domains, a design Morrison and colleagues described in 1984.

A humanized antibody keeps only the six mouse complementarity-determining regions (CDRs), the loops that contact the antigen, grafted into a human framework, as Jones and Winter first did in 1986. Bevacizumab's label describes exactly that: "human framework regions and murine complementarity-determining regions".

A human antibody has no mouse-derived sequence at all. It comes from mice whose antibody genes were replaced with human ones, from phage libraries of human antibody fragments, or from human B cells. Adalimumab, licensed in 2002 and the first human antibody in the Purple Book, was "created using phage display technology", according to its label.

<caption><strong>Table 1. The four antibody origins.</strong> Mouse-derived parts, the INN substem used before 2017, and the count among the 159 FDA-licensed antibody products, classified by ProteinIQ from FDA labels and INNs. Sources: FDA Purple Book (2026), openFDA (2026).</caption>
<thead>
<tr><th>Origin</th><th>From a mouse</th><th>INN substem before 2017</th><th>Example</th><th>FDA-licensed products</th></tr>
</thead>
<tbody>
<tr><td>Murine</td><td>Whole antibody</td><td>-o-</td><td>ibritumomab</td><td>8</td></tr>
<tr><td>Chimeric</td><td>Both variable domains</td><td>-xi-</td><td>rituximab</td><td>14</td></tr>
<tr><td>Humanized</td><td>The six CDR loops, often with a few framework residues</td><td>-zu-</td><td>trastuzumab</td><td>83</td></tr>
<tr><td>Human</td><td>Nothing</td><td>-u-</td><td>adalimumab</td><td>54</td></tr>
</tbody>

Reviews of patient data show the logic of this sequence. Hwang and Foote found that replacing the mouse constant regions cut anti-antibody responses the most, and humanizing the variable domains cut them further.

The gain stops short of zero. When Harding and colleagues mapped helper T-cell epitopes in eight humanized antibodies, every epitope sat in a CDR-containing region, the part that a fully human antibody also has to vary. A human antibody is less likely to provoke a response, not incapable of it.

## How many FDA-approved antibodies are humanized, human, chimeric or murine?

Humanized antibodies are the largest group in every way of counting. Of the 155 antibody products licensed through the end of 2025, 81 are humanized (52%), 52 human (34%), 14 chimeric (9%) and 8 murine (5%). The 150 products that are still licensed are slightly more human, because five of the eight murine products, mostly early imaging and radiolabeled antibodies, have been withdrawn.

<caption><strong>Table 2. FDA-licensed antibody products by origin.</strong> ProteinIQ classification of each antibody product in the FDA Purple Book September 2026 release, counted once at its first license. Percentages are of the column total. Sources: FDA Purple Book (2026), openFDA (2026).</caption>
<thead>
<tr><th>Origin</th><th>Licensed by Sep 25, 2026</th><th>Through 2025</th><th>Still licensed</th></tr>
</thead>
<tbody>
<tr><td>Humanized</td><td>83 (52.2%)</td><td>81 (52.3%)</td><td>81 (54.0%)</td></tr>
<tr><td>Human</td><td>54 (34.0%)</td><td>52 (33.5%)</td><td>53 (35.3%)</td></tr>
<tr><td>Chimeric</td><td>14 (8.8%)</td><td>14 (9.0%)</td><td>13 (8.7%)</td></tr>
<tr><td>Murine</td><td>8 (5.0%)</td><td>8 (5.2%)</td><td>3 (2.0%)</td></tr>
<tr><td>All antibody products</td><td>159</td><td>155</td><td>150</td></tr>
</tbody>

The origins arrived in waves. Murine and chimeric antibodies made up 11 of the first 20 products, licensed through 2004, although humanized antibodies such as daclizumab and trastuzumab had already arrived in 1997 and 1998. Human antibodies appeared in numbers from 2009, when all four antibodies licensed that year were human: golimumab, canakinumab, ustekinumab and ofatumumab.

![Stacked bars of FDA-licensed antibody products by origin in each period: 10 in 1992 to 1999, 10 in 2000 to 2004, 8 in 2005 to 2009, 15 in 2010 to 2014, 46 in 2015 to 2019, 56 in 2020 to 2024 and 14 in 2025 to September 2026, with humanized and human antibodies making up most of every period after 2004](/images/charts/fda-antibody-origin-by-period.webp '**Figure 1. Origin of new FDA-licensed antibodies by period.** ProteinIQ classification of the 159 antibody products in the FDA Purple Book September 2026 release, each counted once in the year of its first license; 2025 to 2026 runs to September 25. Purple Book dates place muromonab-CD3, approved in 1986, in 1992. Sources: [FDA Purple Book (2026)](https://purplebooksearch.fda.gov/downloads), [openFDA (2026)](https://open.fda.gov/apis/drug/label/).')

Murine antibodies have effectively left the field. The last two were blinatumomab in 2014, a bispecific built from two mouse single-chain fragments, and the immunotoxin moxetumomab pasudotox in 2018, since withdrawn. Chimeric antibodies have not. Five were licensed from 2020 to 2024: isatuximab, margetuximab, mirvetuximab soravtansine, ublituximab and zolbetuximab.

## Has the share of fully human antibodies grown?

It grew for two decades and then stopped. Human antibodies were 8% of licensed antibodies at the end of 2002, 23% by 2010 and 33% by 2015. Since then the share has moved between 32% and 37%, and it stood at 34% in September 2026. Over the same years the humanized share climbed from 39% to 52%, while chimeric and murine antibodies shrank to a combined 14%.

![Lines showing the share of all FDA-licensed antibodies by origin from 1998 to 2026: humanized rising to 52%, human rising from zero in 2000 to about 34% after 2015, chimeric falling from 40% to 9% and murine from 30% to 5%](/images/charts/fda-antibody-origin-cumulative-share.webp '**Figure 2. Share of all FDA-licensed antibodies by origin.** ProteinIQ classification of antibody products in the FDA Purple Book September 2026 release; each point is the share of all products licensed by the end of that year, shown for even years, with 2026 to September 25. Sources: [FDA Purple Book (2026)](https://purplebooksearch.fda.gov/downloads), [openFDA (2026)](https://open.fda.gov/apis/drug/label/).')

The flat line comes from new approvals that tipped back toward humanized antibodies. Human antibodies were 41% of products licensed from 2015 to 2019 but 29% from 2020 to 2024, when humanized antibodies took 63%. In 2023 and 2024, 20 of the 25 new antibody products were humanized. The 14 products licensed from January 2025 to September 2026 split evenly, seven human and seven humanized.

Format explains much of the shift. Eleven of the [FDA's 15 antibody-drug conjugates](/guides/fda-approved-adcs) and 11 of its 14 bispecific antibodies are humanized. The two formats made up 20 of the 70 products licensed since 2020, and 16 of those 20 are humanized; among the other 50, human antibodies hold 40%.

Conjugates are often built on antibodies already proven in the clinic: Kadcyla and Enhertu both carry humanized trastuzumab. Across all years, plain monospecific antibodies are nearly even, 56 humanized against 48 human.

<caption><strong>Table 3. Origin of FDA-licensed antibody products by format.</strong> ProteinIQ classification of the 159 antibody products in the FDA Purple Book September 2026 release; conjugates take the origin of their antibody, fragments the origin of their variable domains, and bispecifics the whole-molecule origin on their label. Sources: FDA Purple Book (2026), openFDA (2026).</caption>
<thead>
<tr><th>Format</th><th>Humanized</th><th>Human</th><th>Chimeric</th><th>Murine</th><th>Total</th></tr>
</thead>
<tbody>
<tr><td>Unconjugated monospecific antibodies</td><td>56</td><td>48</td><td>11</td><td>1</td><td>116</td></tr>
<tr><td>Antibody-drug conjugates</td><td>11</td><td>2</td><td>2</td><td>0</td><td>15</td></tr>
<tr><td>Bispecific antibodies</td><td>11</td><td>2</td><td>0</td><td>1</td><td>14</td></tr>
<tr><td>Fragments</td><td>5</td><td>0</td><td>1</td><td>0</td><td>6</td></tr>
<tr><td>Mixtures</td><td>0</td><td>2</td><td>0</td><td>0</td><td>2</td></tr>
<tr><td>Radioconjugates and immunotoxin</td><td>0</td><td>0</td><td>0</td><td>6</td><td>6</td></tr>
<tr><td>All antibody products</td><td>83</td><td>54</td><td>14</td><td>8</td><td>159</td></tr>
</tbody>

Humanization has also become easier to do. Antibodies from immunized mice, rabbits or llamas can still be the best binders against a target, and the step that turns them into drug candidates is now partly computational. [BioPhi](/app/biophi) proposes humanizing mutations and scores how human a sequence looks against millions of antibody sequences from human repertoires, and [Humatch](/app/humatch) picks human acceptor genes while keeping as few non-human residues as possible.

## Why can't you tell an antibody's origin from its name anymore?

For three decades the name gave the answer. A monoclonal antibody INN ended in -mab, and the syllable before it named the source: -ximab for chimeric (rituximab), -zumab for humanized (trastuzumab), -umab for human (adalimumab) and -omab for mouse (ibritumomab). Readers could sort a list of antibodies by origin from their names alone.

Three changes broke that rule. In 2014 the WHO redefined the categories by sequence identity: a variable domain had to be at least 85% identical to human germline sequences to count as human or humanized, or the antibody would be named chimeric. A group of 34 antibody engineers led by Tim Jones called the threshold arbitrary and showed that it would rename established humanized antibodies such as pertuzumab as chimeric.

In 2017 the WHO stopped putting a source syllable in new names at all. In 2022 it replaced -mab for new antibodies with four stems that describe structure rather than species: -tug for unmodified immunoglobulins, -bart for engineered ones, -mig for multispecifics and -ment for fragments.

The effect on the approval list is now large. Of the 159 FDA-licensed antibody products, 41 have names with no source syllable, and that includes 40 of the 70 licensed since January 2020 and 13 of the 14 licensed since January 2025. Cemiplimab, licensed in 2018, was the first. Veligrotug, the first licensed antibody named with -tug, is humanized according to its label, but nothing in its name says so.

The older names can mislead too. Loncastuximab tesirine carries the chimeric -xi- syllable assigned under the 2014 rules, yet its label calls the antibody humanized. Rozanolixizumab carries -xizu-, the 2014 syllable for antibodies with one chimeric and one humanized chain, and its label also says humanized.

For any antibody named after 2017, the label is the most reliable public record. Section 11 of the prescribing information states the origin for 145 of the 159 products, usually in its first sentence. Sequence databases offer a second view: Thera-SAbDab annotates the genetic origin of therapeutic antibody sequences, and [ANARCI](/app/anarci) numbers a variable domain and reports which species' germline genes it matches best.

## Does an antibody's origin still matter?

Origin is a proxy for immunogenicity, and a rough one. Muromonab-CD3 showed what a fully foreign antibody does. Between humanized and human antibodies the line is much fainter, because the residual T-cell epitopes Harding found sit in the CDRs that every antibody, human or not, uses to bind its target. The 2014 naming debate made the same point from the sequence side: somatically mutated human antibodies and well-humanized ones overlap in their identity to human germline genes, so a fixed threshold separates the categories more sharply than biology does.

The practical questions for a new molecule are narrower. [Antibody engineers](/guides/what-is-antibody-engineering) ask which non-human residues remain, whether they create T-cell epitopes, and whether humanization changed the CDR loops that bind the target. Humanness scores from BioPhi answer the first question. Structure predictors such as [ImmuneBuilder](/app/immunebuilder) and [ABodyBuilder3](/app/abodybuilder3) model the variable domains before and after humanization so the loops can be compared, and [TAP2](/app/tap2) flags developability liabilities that origin does not capture.

## Methodology

We started from the per-product list behind our count of FDA-approved monoclonal antibodies: the 159 antibody products in the FDA Purple Book September 2026 release, each counted once at its first license, with Fc fusion proteins, biosimilars and later licenses for the same antibody excluded. We kept its counting rules unchanged, including Purple Book dates, which put muromonab-CD3 in 1992 rather than its 1986 approval.

For each product we read section 11 (Description) of the current FDA label through the openFDA drug labeling API, matched by license number, on October 29, 2026. The label's own word decided the origin for 145 products, including apitegromab, too new for openFDA, whose label we read on DailyMed. Where a label was silent or no longer available, as for nivolumab, blinatumomab, efalizumab and two withdrawn imaging antibodies, we used the INN source syllable, for 10 products in all.

Four products had neither: pozelimab, tarlatamab, pivekimab sunirine and the relatlimab in Opdualag. We classified them from the Thera-SAbDab sequence annotation and, for pivekimab sunirine, from the paper that introduced its humanized antibody.

Conjugates (antibody-drug conjugates, radioconjugates and the immunotoxin) take the origin of their antibody, and fragments, including the humanized nanobody caplacizumab, take the origin of their variable domains. Bispecific antibodies take the whole-molecule origin their label states; tarlatamab, whose label is silent, takes the origin of its less human arm. Both mixtures, Inmazeb and Opdualag, contain only human antibodies. "Human" includes antibodies from human-gene transgenic mice, phage and yeast libraries and human donors, as labels and INNs do.

As a check, we compared our classification with Thera-SAbDab's sequence-based annotation. Of the 141 products it annotates with a single unambiguous origin, 135 agree. The six that differ are all humanized by their label or INN, and the database calls them human (efalizumab, crovalimab), chimeric (loncastuximab, veligrotug) or gives one arm as chimeric (the bispecifics epcoritamab and talquetamab). Twelve products carry Thera-SAbDab's combined "chimeric and/or humanised" annotation, and six old or withdrawn products are absent from it.

The [classification for every product](/data/guides/humanized-vs-fully-human-antibodies/antibody-origins.csv) lists the license number, first license date, format, origin, the evidence used and, for label-based calls, the quoted label sentence and its DailyMed set ID. The [summary](/data/guides/humanized-vs-fully-human-antibodies/summary.json) holds the counts by year, period, format and naming scheme. The analysis script is `scripts/antibody-origin-fda-approvals.py`. We publish our classification and the short label sentences that support it, but not Thera-SAbDab's annotations, which carry no stated licence for redistribution. Both files are free to reuse under CC BY 4.0 with credit to ProteinIQ. We last checked them on October 29, 2026.

ProteinIQ. "Human vs humanized vs chimeric antibodies: the origin of every FDA-approved antibody." Published October 29, 2026. Accessed [your access date]. https://proteiniq.io/guides/humanized-vs-fully-human-antibodies