I have the following models in sqlalchemy:

class Location(Base):
  __tablename__ = "locations"

  predictions: Mapped[list[Prediction]] = relationship(
    back_populates="location",
    foreign_keys="Prediction.location_id",
  )


prediction_relations = Table(
    "prediction_relations",
    Base.metadata,
    Column("dependant_prediction_id", ForeignKey("predictions.id"), primary_key=True),
    Column("underlying_prediction_id", ForeignKey("predictions.id"), primary_key=True),
)


class Prediction(Base, UUIDMixin):
    __tablename__ = "predictions"

    location_id: Mapped[UUID] = mapped_column(ForeignKey("locations.id"))
    location: Mapped[Location] = relationship(
        back_populates="predictions", foreign_keys=[location_id]
    )
    underlying_predictions: Mapped[list["Prediction"]] = relationship(
        "Prediction",
        secondary=prediction_relations,
        primaryjoin="Prediction.id == prediction_relations.c.dependant_prediction_id",
        secondaryjoin="Prediction.id == prediction_relations.c.underlying_prediction_id",
        back_populates="dependant_predictions",
    )
    dependant_predictions: Mapped[list["Prediction"]] = relationship(
        "Prediction",
        secondary=prediction_relations,
        primaryjoin="Prediction.id == prediction_relations.c.underlying_prediction_id",
        secondaryjoin="Prediction.id == prediction_relations.c.dependant_prediction_id",
        back_populates="underlying_predictions",
    )

When I try to add a prediction that has an underlying prediction I get the following error:
sqlalchemy.exc.IntegrityError: (sqlite3.IntegrityError) NOT NULL constraint failed: predictions.location_id.

Is it possible to store them both together or do I have to commit the underlying prediction first before the dependant prediction can be commited?