The committee on recombinant DNA research of the National Institutes of Health recommended yesterday that the federal government withdraw from direct supervision of up to 97 percent of the DNA research being carried on in the United States.

The recommendation, if approved by the head of NIH, would mean that the task of registering and reviewing 97 percent of the recombinant work would go to hundreds of "biosafety committees" established at university, industry and other research laboratories around the nation.

Since last January, when the panel voted to get NIH out of the business of routinely supervising research guidelines it established in 1976, these committees have been overseeing the safest 80 percent of DNA research. Yesterday's vote recommends to the NIH director that about 17 percent more be placed under local supervision. The only DNA research that would remain under government supervision are unusual cases and those not easily grouped according to their level of hazard.

Since the 1976 NIH guidelines give detailed instructions for researchers, the chief job for the local biosafety groups is simply to place proposed research in the proper category of estimated risk. Each category calls for a particular level of safety precautions -- from a few simple cleanliness procedures to whole rooms that must be specially sealed and ventilated.

Recombinant DNA, also called gene splicing, puts alien genes into small organisms, giving them new characteristics -- such as the ability to make human hormones. Some people fear that the organisms may develop unforeseen and dangerous properties, and then escape the laboratories to cause disease.

Patricia King, a Justice Department lawyer and a member of the committee, said during debate that the "only question is whether the IBCs [local Institutional Biosafety Committees] can do as good a job in categorizing these things as our committee."

In a test done in 1978, it was found that local biosafety committees made mistakes in categorizing experiments only 5 percent of the time, and some of those were mistakes of form-filling rather than substance, according to Maxine Singer of the National Cancer Institute, who brought the proposal before the NIH board yesterday.

In a straw vote yesterday, the NIH recombinant committee also signaled its intention to continue, in special industrial cases, supervising both the organisms and the containment hardware. Most major companies doing the research have voluntarily submitted their work to the NIH committee for review even though the law does not require them to.

The straw vote indicated a reversal of mood in the committee since its previous meeting. During that meeting, some members suggested giving up supervision of at least the hardware part of the industrial work. The committee, they said, has expertise in biologicial matters, but not in industrial hardware matters. In moving recombinant research from the level of the test tube to the level of the multigallon vat, a whole new level of equipment is introduced.

The committee found itself in the position of "just accepting the word of the industry on the safety of this equipment, instead of actually giving it a critical view," said Patricia King.

But between meetings, the NIH committee received letters from the Occupational Safety and Health Administration and others suggesting that no other agency with the proper expertise could or would start reviewing those matters if NIH dropped them.

So committee member Sheldon Krimsky of Tufts University drew up a motion, approved in the straw vote, that would pass responsibility for matters of industrial hardware inot a subcommittee at NIH that would comprise members from several other government agencies -- including OSHA. This subcommittee would then pass on industrial hardware issues after the full committee made its biological recommendations on the industries' research.