Plate data lands and the analysis runs

The last selection round handed back a panel of candidate antibodies, and the binding ELISA that ranks them just came off the plate reader. Each candidate is titrated across an eight-point dilution series, with control wells for background and a positive control. Ranking the panel means turning each candidate's dilution series into a dose-response curve and reading its EC50, with every EC50 tied back to the antibody it came from.

The reader's export drops into Affinity as structured results. Its file format is recognized and parsed into a table, so the optical densities arrive already mapped to their wells and tied to the candidates loaded on the plate. No script to program and run, no manual data entry to perform, no separate step to say which well held which antibody.

The scientist applies a plate template to label the wells against the plate map, delineating candidate wells from controls, and Affinity works the plate from there: replicates averaged, signal normalized against the controls, and each candidate's series fit to a dose- response curve, with the EC50 falling out of each fit and the curves that misbehaved flagged rather than buried. Every step writes a new dataset instead of overwriting the last, so trying a tighter normalization or masking a stray point branches off the original rather than destroying it. The plates reads become a column of comparable potencies attached to the right molecules, and the decision of which antibodies to advance picks up from there.

You annotate and screen an antibody's sequence

A candidate antibody's sequence arrives, and before anyone commits expression slots to it, there are two questions. What is it: which V, D, and J germlines it came from, and where the framework and CDR boundaries fall. And will it cause trouble later: the deamidation, oxidation, glycosylation, and charge motifs that sink an otherwise good binder partway through development.

Registering the antibody or sequencing a plate of clones answers both as the sequence data lands. Affinity runs an IgBLAST-based analysis against the germlines you've registered, identifying the variable regions and delineating and numbering their frameworks and CDRs, so that sequences enter the registry as structured information, paired domains with their framework and CDR breakdown. In the same pass, it screens those regions against a set of known (or your custom) sequence-liability patterns and flags what it finds: an N-glycosylation site, a deamidation-prone region, or an unpaired cysteine, each marked by position.

What lands on a clone or antibody's page is a fully annotated variable domain with its liabilities called out, and because those liability counts and CDR properties are the same sequence-derived fields a scoring profile reads, a developability red flag counts against the candidate automatically when the panel is later ranked. The analysis is performed when the sequence is first registered, not as a separate step that someone has to remember to run later.

You identify clones from an NGS run

A repertoire run came back from the sequencer, and the pipeline that annotates it has already done its work, with V(D)J calls made, CDRs delineated, productivity scored, and the whole run written out in standard AIRR format. What's left is the part that decides the outcome: turning a few hundred thousand annotated clonotypes into the handful of clones worth expressing.

That AIRR output is what comes into Affinity. Create an analysis, upload the files, and Affinity organizes the annotated repertoire and makes the run navigable: enrichment, mutation frequency, CDR properties, and built-in sequence alignment to compare across clonotypes. From there, the clone picker is how the scientist builds a working set: narrow to a list of clones, identify the candidates worth taking forward, add them to a set, align them, or send them straight to expression.

You search and align across the registry

Program review is Friday. The work to assemble: every assay result, every called variable domain, every score, every liability flag for the candidates in the top set, all composed into a single view the team can read together rather than a deck rebuilt the night before by exporting from each surface.

Affinity's report builder assembles exactly that. The scientist opens it from any list of candidates — a saved set, a Lots page, the project's dashboard — and the candidates already in scope arrive pre-filtered. Pick the result columns to include (KD from Biacore, EC50 from the ELISA, monomer percent from SEC, Tm from nanoDSF), add the sequence-derived properties (CDR lengths, liability counts, computed protein properties), and the builder composes one row per candidate with every field side by side, sortable by any column and ready to export for the deck or to stay open as the live working surface through the meeting. A companion alignment view runs MUSCLE over the same set's chains, so the structural picture sits one tab away from the numerical one.

When the report needs to outlive one scientist's session, it is saved as a named report with its columns and criteria preserved, ready to re-open the next time a similar review comes around.

You search the registry for a sequence

A reviewer asks whether a CDR3 you're looking at has come up before, perhaps in an earlier campaign, a different program, or a deposited reference. The work is a lookup across everything the team has registered, not a guess from memory.

Sequence search is built in: DNA or protein, by exact match or by BLAST, against every registered protein, plasmid, oligonucleotide, and clone, each hit linking back to its source. Alignment works the same way, with sequences picked from the registry by name or pasted as FASTA, aligned by the chosen method, colored by residue against a highlighted reference, and downloaded as aligned FASTA or a rendered image. Translation renders DNA and protein side by side, with FASTA or CSV export. The bioinformatics is part of the platform, not a separate license.