The Signal
We analyze SEC filings using machine learning classifiers that detect specific language patterns. Raw scores are noisy — they bounce around from filing to filing. To cut through the noise, we compare each filing to the company's recent history and ask: is this filing unusually strong or weak?
above ATYR's recent average — a rare event.
What Does This Mean? (Plain English)
Think of it like a student's test scores. If a student usually scores between 70-90, and suddenly scores 140, that's unusual. We measure "how unusual" using standard deviations:
ATYR's Q3 2025 filing scored +3.4σ — the highest in their 10-year filing history. This isn't noise; something changed in how management is describing their business.
The Pattern
Looking at the chart, you can see ATYR's filing strength has been below average (below the zero line) for most of 2024 and 2025. This declining trend suddenly reversed in Q3 2025 with a massive spike.
This pattern — prolonged weakness followed by a sudden spike — often indicates a material change in the company's outlook. Management language in SEC filings tends to become more forward-looking and confident when catalysts approach.
What Triggered the Spike?
The Q3 2025 filing showed unusual strength in three specific areas:
FDA Aligned (+168%): Language about productive FDA interactions, clear regulatory pathway, or alignment on trial design.
Clinical Progress (+135%): Language about enrollment, dosing, or trial advancement.
How We Calculate This
We run 9 biotech-specific ML classifiers on each SEC filing to create a "biotech momentum" score. Then we compare each filing to the prior 8 quarters to calculate a rolling z-score:
Z-Score = (Current Score − Average of Last 8 Quarters) ÷ Standard Deviation
A z-score of +3.4 means the Q3 2025 filing scored 3.4 standard deviations above ATYR's recent average. Given the historical distribution, there's less than a 0.1% chance this happened by random variation.