Real-Time Congress Intelligence for Medical Affairs Teams

It's the Friday before ASCO. Your MSL team is about to be on the floor in Chicago, and the abstract titles dropped overnight. Somewhere in those few thousand abstracts are the three that matter for your asset, two that matter for the competitor everyone keeps asking about, and one late-breaker that's going to reframe the whole subtype conversation. The question every medical affairs lead asks at that moment is the same: who's reading all of this, and how fast.

The honest answer at most mid-size biotechs is one or two people, a shared spreadsheet, and a weekend. That doesn't scale, and it doesn't survive the four or five congresses that actually move oncology each year. There's a better primitive than manual triage, and it starts with the fact that congress abstracts now carry machine-readable identifiers.

How can medical affairs teams monitor congress abstracts in real time?

Major oncology congresses register their abstracts as DOIs in journal supplements, so a team can pull new abstract metadata (title, authors, affiliations) by polling Crossref for the supplement issue as it publishes, then filter that stream against the team's own gene, drug, and competitor watchlist. No manual scraping, no waiting for a post-congress report.

The mechanism is simpler than it sounds. ASCO publishes its abstracts in the Journal of Clinical Oncology annual meeting supplement (ISSN 1527-7755). AACR publishes in Cancer Research supplements (ISSN 1538-7445). ESMO uses Annals of Oncology, ASH uses Blood, and the lung crowd has WCLC through the Journal of Thoracic Oncology. Each abstract gets a DOI, and each DOI lands in Crossref with structured metadata: title, author list, author affiliations, and the supplement it belongs to.

That means you can ask Crossref a precise question: give me every record in Cancer Research volume X, supplement 1, published in the last 24 hours. The answer comes back as JSON. You don't need the publisher's full text, and you shouldn't take it. The metadata layer (titles, authors, affiliations, abstract numbers) is enough to know what was presented, by whom, and from which institution. The clinical reading still happens with a human, on the abstract itself, through the congress's own channels.

UNMIRI's Engine 3 pulls this Crossref-indexed congress metadata on a continuous schedule and matches each new record against a structured oncology graph. When an abstract title mentions a gene, a drug, or a trial you're watching, it surfaces. When it names an author your team has flagged as a key opinion leader, it surfaces. The triage that used to eat a weekend becomes a filtered feed.

What does "real-time" actually mean for congress data?

Real-time here means abstract metadata appears in your feed within hours of the supplement going live in Crossref, not days after a vendor compiles a recap deck. The lag is the indexing lag, which for the major oncology journals is short and getting shorter.

There's a meaningful difference between three timelines. The first is the post-congress recap: a polished summary that lands a week or two after the meeting, useful for archives, useless for the hallway conversation happening right now. The second is the live-tweet-and-LinkedIn timeline, which is fast but noisy, unstructured, and impossible to map cleanly to your asset. The third is the metadata timeline: the moment an abstract gets a DOI and that DOI shows up in Crossref's index.

That third timeline is the one worth building on. It's structured, it's attributable, and it's tied to the canonical record rather than someone's paraphrase. For a medical affairs team, the practical payoff is a congress-prep workflow that updates itself. Your watchlist for, say, KRAS G12C in non-small cell lung cancer doesn't wait for the recap. The moment a relevant abstract is indexed, it's in your queue with the author affiliations attached, ready for an MSL to read before the session.

What should an oncology medical affairs team track across congresses?

The high-value signals are competitor data readouts, new mechanism or biomarker abstracts in your tumor type, late-breakers, and the institutional and author footprint behind each, all mapped to your existing watchlists so nothing relevant gets lost in the volume.

A few concrete patterns from how oncology medical affairs teams actually work:

• Competitive readouts. If a competitor has a phase III in your indication, you want the abstract the moment it's indexed, not when the press release frames it. The metadata tells you the trial, the presenting author, and the institution. The reading tells you the rest.
• Biomarker and subtype shifts. A new abstract on HER2-low breast cancer, or on a resistance mechanism like EGFR C797S after osimertinib, can change the questions MSLs get in the field within days. Catching it early means your medical team is ahead of the curve instead of reacting to it.
• KOL activity. Congresses are where influence is visible. Which authors are presenting, from which institutions, on which mechanisms. That maps directly into pre-call and advisory-board planning, which is a whole discipline of its own (covered in the post on mapping oncology KOLs from open data).
• Late-breakers. The ones that aren't in the main abstract book until the last minute. A continuous poll catches them; a one-time pre-congress export doesn't.

The volume is the problem and the structure is the solution. ASCO alone runs several thousand abstracts. No team reads all of them. A team that has pre-defined its genes, drugs, tumor types, competitors, and watched authors reads the forty that matter and ignores the rest with confidence.

How is congress intelligence different from a literature alert?

A literature alert tells you when a paper is published, which for oncology often means months after the data was first shown at a congress. Congress intelligence closes that gap by tracking the abstract at the moment of presentation, where most practice-shaping oncology data debuts first.

Oncology runs on congresses. The pivotal readout that becomes a New England Journal of Medicine paper in the fall was usually a plenary abstract at ASCO in June. If your surveillance only watches PubMed and Europe PMC, you're structurally late on the data that matters most, because the journal publication trails the presentation by a full quarter or more.

That's why congress intelligence and literature surveillance are complementary, not redundant. The literature feed gives you the durable, full-text, citable record. The congress feed gives you the early signal. UNMIRI runs both on the same evidence graph, so an abstract caught at ASCO in June links forward to the JCO or NEJM paper when it lands, and the provenance chain is intact the whole way. You see the trajectory of a piece of evidence, not just its endpoints.

Where does AI fit, and where does it not?

AI is useful for matching, ranking, and clustering the abstract stream against your watchlists. It should not be inventing what an abstract says. The metadata is the ground truth, and the clinical interpretation stays with the human reading the source.

This is the line that matters for a defensible medical affairs function. The match logic (does this abstract title touch a gene, drug, trial, or author I care about) is a structured-graph problem, not a generative one. When UNMIRI ranks a congress abstract into your feed, it's because the title and metadata matched a node in the graph you told it to watch, not because a model guessed at relevance.

Any summarization that does happen is grounded in the actual metadata, carries the DOI and source attribution, and is built for a human to review before it informs a single field interaction. That's the same posture that governs the citation-grounded literature Q&A. Medical affairs operates above the patient layer (no PHI in any of these queries, by design) but the standard for evidentiary discipline is just as high, because a wrong claim about a competitor's data or a misattributed result is its own kind of failure.

The architecture is GraphRAG over a Neo4j evidence graph carrying CIViC and ClinVar citations, layered with live PubMed, Europe PMC, ClinicalTrials.gov, and the Crossref-indexed congress feed. Every signal traces back to a source. That's the part that makes a congress-intelligence workflow something a regulatory-minded organization can actually adopt.

Putting it together

A medical affairs team that wants to stop losing weekends to congress triage needs three things: a structured watchlist of its genes, drugs, tumor types, competitors, and KOLs; a feed that polls the canonical metadata sources as they publish; and a discipline of grounding every surfaced signal back to its source. The congresses already register their abstracts as DOIs. The metadata is already in Crossref. The work is connecting that stream to what your team actually cares about and ranking it so the forty abstracts that matter rise above the several thousand that don't.

That's what Engine 3, UNMIRI's literature-intelligence product, is built to do. It's in early access, not general availability, and we're honest about that distinction throughout. If you run medical affairs at an oncology biotech and the next congress is already on your calendar, this is the workflow worth getting ahead of.

Umair Khan

Founder and CTO, UNMIRI

Building UNMIRI, a precision oncology infrastructure company with four product surfaces: cross-vendor NGS interpretation, genomics-aware decision support, oncology literature intelligence, and a free cross-vendor unification tool for clinicians. Writing here on architecture, clinical data, and HIPAA-ready AI.

See the NGS Interpretation API →

Clinical advisors: UNMIRI is in active conversations with multiple board-certified pathologists about formal advisory roles. Public introductions land on the About page once each engagement is formalized and the advisor approves being named.